A First Course in Object Oriented

Development
A Hands-On Approach

Leif Lindbäck

May 17, 2017

Revision History

Date Description Page (at

time of
revision)
2016-02-25 First published version. Sections 6.6-6.10 and chapter 7 All
are not yet written.
2016-04-01 Fixed typos Section 5.6
2016-04-05 Class Amount was kept, by mistake, after it had been 29-31
stated that it should be removed.
2016-04-05 Small changes to make the text clearer. 69-71
2016-04-05 Improved table layout 26
2016-04-15 Create stereotype was missing in some classes in design 64, 68, 71,
class diagrams. 75, 77
2016-04-15 Added section 6.6 103-118
2016-04-15 Added section 6.7 118-119
2016-04-15 Moved testing to a separate chapter 120
2016-04-15 Improved layout of table 6.1 81
2016-06-02 Clarified description of naïv domain model. 31
2016-06-07 Added chapter seven. 130-149
2016-06-07 Split chapter eight. 150-153
2016-10-25 Added exceptions and association name to UML cheat 159-163
sheet.
2016-12-06 Added chapter eight. 150-174
2016-12-06 Changed code reference to GitHub repository. 280
2016-12-19 Changed license. ii
2017-01-27 Added chapter nine (not yet completed) 176-205

i
 Revision History

Date Description Page (at

time of
revision)
2017-01-27 Removed the Further Reading sections 2, 14, 21, 40,
81, 150
2017-01-27 Removed references to specific courses from preface iii
2017-01-31 Added singleton pattern to section 9.4 204-207
2017-02-23 Added section 9.3 183-192
2017-03-09 Added template method pattern 215-220
2017-03-10 Added warning not to model parts as attributes in 34-35
domain model.
2017-03-14 Changed legacy time to java.util.time 91, 92, 117,
142, 143, 145
2017-03-20 Fixed typos 17, 21
2017-04-11 Added composite pattern 214-218
2017-04-27 Wrong caption to figure 7.16 143
2017-04-27 Wrong code in figure 7.17 144
2017-05-17 Attributes were, by mistake, not private in listing 7.17 144

ii
License
Except for figures 5.7, 5.8, 5.11, 5.14, 7.1, 7.2, 7.6, 8.1, 9.4, 9.11, 9.13, and 9.18 A First Course
in Object Oriented Development, A Hands-On Approach by Leif Lindbäck is licensed under a
Creative Commons Attribution 4.0 International License, see http://creativecommons.
org/licenses/by/4.0/.

iii
Preface
This text is a compilation of material developed during fifteen years of teaching a first course
on object oriented development. Students taking that course have previously taken one 7.5
hp credit course in object-oriented programming. Thus, the reader is assumed to have basic
knowledge of Java programming. Important concepts, in particular objects and references, are
repeated in chapter 1.
Things that are crucial to remember, but easy to miss, are marked with an exclamation
mark, like this paragraph. Forgetting the information in such paragraphs might lead to severe !
misunderstandings.
Paragraphs warning for typical mistakes are marked NO!, like this paragraph. Such para-
graphs warn about mistakes students have frequently made.
NO!
There are Java implementations of all UML design diagrams in Appendix C. The purpose
is to make clearer what the diagrams actually mean. The analysis diagrams can not be imple-
mented in code, since they do not represent programs. There is a NetBeans project with all
Java code in this book, that can be downloaded from GitHub [Code].

iv
Contents
Revision History . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . i

License . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iii

Preface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iv

I Background 1

1 Java Essentials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.1 Further Reading . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.2 Objects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.3 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.4 Arrays and Lists . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.5 Exceptions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
1.6 Javadoc . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
1.7 Annotations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
1.8 Interfaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
1.9 Inheritance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10

2 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
2.1 Why Bother About Object Oriented Design? . . . . . . . . . . . . . . . . . . 13
2.2 Software Development Methodologies . . . . . . . . . . . . . . . . . . . . . 13
2.3 Activities During an Iteration . . . . . . . . . . . . . . . . . . . . . . . . . . 14
2.4 Unified Modeling Language, UML . . . . . . . . . . . . . . . . . . . . . . . 15

3 The Case Study . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17

3.1 Basic Flow . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
3.2 Alternative Flows . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18

II Course Content 19

4 Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
4.1 UML . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
4.2 Domain Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
4.3 System Sequence Diagram . . . . . . . . . . . . . . . . . . . . . . . . . . . 34

v
 Contents

5 Design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
5.1 UML . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
5.2 Design Concepts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
5.3 Architecture . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
5.4 A Design Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
5.5 Designing the RentCar Case Study . . . . . . . . . . . . . . . . . . . . . . 58
5.6 Common Mistakes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78

6 Programming . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
6.1 Dividing the Code in Packages . . . . . . . . . . . . . . . . . . . . . . . . . 80
6.2 Code Conventions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
6.3 Comments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
6.4 Code Smell and Refactoring . . . . . . . . . . . . . . . . . . . . . . . . . . 83
6.5 Coding Case Study . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104
6.6 Common Mistakes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118

7 Testing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 120
7.1 Unit Tests and The JUnit Framework . . . . . . . . . . . . . . . . . . . . . . 121
7.2 Unit Testing Best Practices . . . . . . . . . . . . . . . . . . . . . . . . . . . 126
7.3 When Testing is Difficult . . . . . . . . . . . . . . . . . . . . . . . . . . . . 128
7.4 Unit Testing Case Study . . . . . . . . . . . . . . . . . . . . . . . . . . . . 129
7.5 Common Mistakes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 148

8 Handling Failure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 149

8.1 UML . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 149
8.2 Exception Handling Best Practices . . . . . . . . . . . . . . . . . . . . . . . 151
8.3 Complete Exception Handling in the Case Study . . . . . . . . . . . . . . . . 172
8.4 Common Mistakes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 173

9 Polymorphism and Inheritance . . . . . . . . . . . . . . . . . . . . . . . . . . . 174

9.1 UML . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 174
9.2 Polymorphism . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 175
9.3 Inheritance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 182
9.4 Gang of Four Design Patterns . . . . . . . . . . . . . . . . . . . . . . . . . . 192

III Appendices 224

A English-Swedish Dictionary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 225

B UML Cheat Sheet . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 227

C Implementations of UML Diagrams . . . . . . . . . . . . . . . . . . . . . . . . . 232

Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 346

vi
 Contents

Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 347

vii
 Part I

Background

1
Chapter 1

Java Essentials
This text assumes previous knowledge of an object oriented programming language. All code
listings are written in Java. This chapter repeats important concepts of Java, but does not cover
the whole language.

1.1 Further Reading

The code examples are available in a NetBeans project, which can be downloaded from
GitHub [Code]. Video lectures, which, in addition to explaining concepts, give an introduction
to NetBeans, are available on the course web [CW].

1.2 Objects
The goal of object-oriented programming is to declare classes, which are groups of data and
methods operating on that data. A class represents an abstraction, for example person. An
object represents a specific instance of that abstraction, for example the person you. Whenever
a new person shall be represented in the program, a new object of the Person class is created.
This is done with the operator new, as illustrated on lines one and five in listing 1.1.

1 Person alice = new Person("Main Street 2");

2 System.out.println("Alice lives at " +
3 alice.getHomeAddress());
4
5 Person bob = new Person("Main Street 1");
6 System.out.println("Bob lives at " + bob.getHomeAddress());
7
8 alice.move("Main Street 3");
9 System.out.println("Alice now lives at " +
10 alice.getHomeAddress());
11 System.out.println("Bob still lives at " +
12 bob.getHomeAddress());

Listing 1.1 Creating and calling objects.

2
 Chapter 1 Java Essentials

Two different objects, representing the persons Alice and Bob, are created in listing 1.1.
Note that when Alice moves to another address, line eight, Bobs address remains unchanged,
since alice and bob are different objects. The output when running the program is provided
in listing 1.2, and the source code for the Person class is in listing 1.3.

1 Alice lives at Main Street 2

2 Bob lives at Main Street 1
3 Alice now lives at Main Street 3
4 Bob still lives at Main Street 1

Listing 1.2 Output of program execution

1 public class Person {

2 private String homeAddress;
3
4 public Person() {
5 this(null);
6 }
7
8 public Person(String homeAddress) {
9 this.homeAddress = homeAddress;
10 }
11
12 public String getHomeAddress() {
13 return this.homeAddress;
14 }
15
16 public void move(String newAddress) {
17 this.homeAddress = newAddress;
18 }
19 }

Listing 1.3 The Person class

A constructor is used to provide initial values to an object. In listing 1.3, the value passed to
the constructor is saved in the object’s field on line nine. Sending parameters to a constructor
is just like sending parameters to a method. More than one constructor is needed if it shall
be possible to provide different sets of initialization parameters. The constructor on lines four
to six is used if no home address is specified when the object is created, the constructor on
lines eight to ten is used when a home address is specified. Note that, on line five, the first
constructor calls the second constructor, using null as the value of the home address.
The variable this always refers to the current object. The variable this.homeAddress
on line nine in listing 1.3 is the field declared on line two, homeAddress on line nine is the
constructor parameter homeAddress declared on line eight. These two are different variables.
A word of warning: use static fields and methods very restrictively! Static fields are shared

3
 Chapter 1 Java Essentials

by all objects of the class. If for example the person’s address was static, all persons would
have the same address. Such a program would be useless. Since fields can not be static, neither
can methods since static methods can access only static fields. Static fields and methods are
normally not used at all, except in a few, very special, cases.

1.3 References
When the new operator is used to create an object, it returns a reference to that object. A
reference can, like any other value, be stored in variables, sent to methods, sent to constructors,
etc. This is illustrated in listing 1.4, which contains a program where a person places food in a
dog’s bowl. First, a Bowl object is created on line three. The reference to that object is stored
in the bowl variable and passed to the constructor of a Person object on line four. On line
16, the Person object stores the reference in the bowl field, declared on line 12. Then, on
line six, the main method calls the feedDog method in person. In feedDog, the method
addFood is called in the previously created bowl, on line 20. This shows how an object
(bowl) can be created in one place (main), passed to another object (person) and used there.

1 public class Startup {

2 public static void main(String[] args) {
3 Bowl bowl = new Bowl();
4 Person person = new Person(bowl);
5 for (int i = 0; i < 9; i++) {
6 person.feedDog();
7 }
8 }
9 }
10
11 public class Person {
12 private Bowl bowl;
13 private int gramsToAdd = 200;
14
15 public Person(Bowl bowl) {
16 this.bowl = bowl;
17 }
18
19 public void feedDog() {
20 bowl.addFood(gramsToAdd);
21 }
22 }
23
24 public class Bowl {
25 private int gramsOfFoodInBowl;
26
27 public void addFood(int grams) throws Exception {
28 gramsOfFoodInBowl = gramsOfFoodInBowl + grams;

4
 Chapter 1 Java Essentials

29 }
30 }

Listing 1.4 The bowl object is created in the main method and a reference to it is
passed to the person object, where it is used.

1.4 Arrays and Lists

An ordered collection of elements can be represented both as the language construct array
and as an instance of java.util.List. An array is appropriate if the number of elements is
both fixed and known, see listing 1.5, where there are exactly five elements.

1 int[] myArray = new int[5];

Listing 1.5 An array has an exact

number of elements, five in this
case.

It is better to use a java.util.List if the number of elements is not both fixed and
known, see listing 1.6.

1 import java.util.ArrayList;
2 import java.util.List;
3 ...
4 List myList = new ArrayList();
5 myList.add("Hej");
6 myList.add(3);

Listing 1.6 Any number of elements can

be added to a List.

A List can contain objects of any class, listing 1.6 stores a String on line five and an
Integer on line six. This means that when reading from the List, the type of the read
element will always be java.lang.Object. It is up to the programmer to know the actual
type of the element and cast it to that type. This procedure is error-prone, it is better to
restrict list elements to be objects of one specific type. This is done in listing 1.7, where
adding <String> on line four specifies that the list may contain only objects of type String.
Adding <> specifies that this holds also for the created ArrayList.

1 import java.util.ArrayList;
2 import java.util.List;
3 ...
4 List<String> myList = new ArrayList<>();
5 myList.add("Hej");
6 myList.add("Hopp");

5
 Chapter 1 Java Essentials

Listing 1.7 A List allowed to contain only objects of type

String.

When list content is restricted to one type, it is possible to iterate the list using a for-each
loop, see lines eight to ten in listing 1.8.

1 import java.util.ArrayList;
2 import java.util.List;
3 ...
4 List<String> myList = new ArrayList<>();
5 myList.add("Hej");
6 myList.add("Hopp");
7
8 for(String value : myList) {
9 System.out.println(value);
10 }

Listing 1.8 Iterating a List with a for-each loop

1.5 Exceptions
Exceptions are used to report errors. When an exception is thrown, the method throwing it
is immediately interrupted. Execution is resumed in the nearest calling method with a try
block. This is illustrated in listing 1.9. On line 34, the addFood method checks if the bowl
would become overfull when more food is added. If so, instead of adding food, it throws an
exception (lines 35-40). This means line 42 is not executed. Instead, the program returns to
the calling statement, which is on line 24, in the feedDog method. However, that line is not
in a try block, which means execution returns to the statement where feedDog was called.
That call is made on line eight, which is in a try block. Execution then jumps immediately
to the corresponding catch block. This means line eleven is the line executed immediately
after throwing the exception on line 35.

1 public class Startup {

2 public static void main(String[] args) {
3 Bowl bowl = new Bowl();
4 Person person = new Person(bowl);
5 try {
6 for (int i = 0; i < 9; i++) {
7 System.out.println("Feeding dog");
8 person.feedDog();
9 }
10 } catch (Exception e) {
11 e.printStackTrace();

6
 Chapter 1 Java Essentials

12 }
13 }
14 }

15 public class Person {

16 private Bowl bowl;
17 private int gramsToAdd = 200;
18
19 public Person(Bowl bowl) {
20 this.bowl = bowl;
21 }
22
23 public void feedDog() throws Exception {
24 bowl.addFood(gramsToAdd);
25 }
26
27 }
28
29 public class Bowl {
30 private int gramsOfFoodInBowl;
31 private static final int MAX_CAPACITY = 500;
32
33 public void addFood(int grams) throws Exception {
34 if (gramsOfFoodInBowl + grams > MAX_CAPACITY) {
35 throw new Exception("Bowl is overfull. " +
36 "Trying to add " +
37 grams + " grams of " +
38 food when there are " +
39 gramsOfFoodInBowl +
40 " grams in bowl.");
41 }
42 gramsOfFoodInBowl = gramsOfFoodInBowl + grams;
43 }
44 }

Listing 1.9 An exception is thrown if food is added to the bowl when it is already full.

All methods in listing 1.9 that may throw an exception declare that, with throws Exception
in the method declaration. This is required if the thrown exception is a checked exception, but
not if it is a runtime exception. An exception is a runtime exceptions if it inherits the class
java.lang.RuntimeException.

7
 Chapter 1 Java Essentials

1.6 Javadoc
Javadoc is used to generate html pages with code documentation, like the documentation of
the Java APIs at http://docs.oracle.com/javase/8/docs/api/. It is strongly rec-
ommended to write Javadoc for all declarations (classes, methods, fields, etc) that are not
private. A Javadoc comment is written between /** and */. The tags @param and @return
are used to document method parameters and return values. See listing 1.10 for examples.

1 /**
2 * A person that lives at the specified address.
3 */
4 public class Person {
5 private String homeAddress;
6
7 /**
8 * Creates a new <code>Person</code>.
9 */
10 public Person() {
11 this(null);
12 }
13
14 /**
15 * Creates a new <code>Person</code> that lives at the
16 * specified address.
17 *
18 * @param homeAddress The newly created
19 * <code>Person</code>’s home address.
20 */
21 public Person(String homeAddress) {
22 this.homeAddress = homeAddress;
23 }
24
25 /**
26 * @return The <code>Person</code>’s home address.
27 */
28 public String getHomeAddress() {
29 return this.homeAddress;
30 }
31
32 /**
33 * The the <code>Person</code> moves to the specified
34 * address.
35 *
36 * @param newAddress The <code>Person</code>’s new
37 * home address.
38 */

8
 Chapter 1 Java Essentials

39 public void move(String newAddress) {

40 this.homeAddress = newAddress;
41 }
42 }

Listing 1.10 Class with javadoc comments

1.7 Annotations
Annotations are code statements that are not executed. Instead, they provide information
about a piece of source code for the compiler, JVM or something else. Annotations are usu-
ally used for properties unrelated to the functionality of the source code, for example to con-
figure security, networking or tests. An annotation starts with the at sign, @, for example
@SomeAnnotation. Annotations may take parameters, for example @SomeAnnotation(
someString = "abc"). An example is found on line 20 in listing 1.11.

1.8 Interfaces
An interface is a contract, specified in the form of method declarations. A class implementing
the interface must fulfill the contract, by providing implementations of the methods. The
method implementations in the implementing class must do what is intended in the method
declarations in the interface. This should be documented in javadoc comments. Note that
the interface contains only declarations of methods, there are no method bodies. Listing 1.11
shows an interface that defines the contract Print the specified message to the log, lines one to
eleven. It also shows a class that implements the interface and fulfills the contract, lines 12-24.

1 /**
2 * An object that can print to a log.
3 */
4 public interface Logger {
5 /**
6 * The specified message is printed to the log.
7 * @param message The message that will be logged.
8 */
9 void log(String message);
10 }
11
12 /**
13 * Prints log messages to <code>System.out</code>.
14 */
15 public class ConsoleLogger implements Logger {
16 /**
17 * Prints the specified string to <code>System.out</code>.

9
 Chapter 1 Java Essentials

18 * @param message The string that will be printed.

19 */
20 @Override
21 public void log(String message) {
22 System.out.println(message);
23 }
24 }

Listing 1.11 Interface and implementing class

The @Override annotation on line 20 in listing 1.11 specifies that the annotated method
should be inherited from a superclass or interface. Compilation will fail if the method is
not inherited. Always use @Override for inherited methods since it eliminates the risk of
accidentally specifying a new method, for example accidentally naming the method logg
instead of log in the implementing class in listing 1.11.

1.9 Inheritance
When a class inherits another class, everything in the inherited class that is not private becomes
a part also of the inheriting class. The inherited class is often called superclass and the inherit-
ing class is called subclass. This is illustrated in listing 1.12, where methodInSuperclass
is declared in Superclass on line two, but called on line eleven as if it was a member of
Subclass. Actually, it has become a member also of Subclass, because it has been in-
herited.

1 public class Superclass {

2 public void methodInSuperclass() {
3 System.out.println(
4 "Printed from methodInSuperclass");
5 }
6 }
7
8 public class Subclass extends Superclass {
9 public static void main(String[] args) {
10 Subclass subclass = new Subclass();
11 subclass.methodInSuperclass();
12 }
13 }

Listing 1.12 methodInSuperclass exists also in the inher-

iting class, Subclass.

A method in the subclass with the same signature as a method in the superclass will override
(omdefiniera) the superclass’ method. This means that the overriding method will be executed
instead of the overridden. A method’s signature consists of its name and parameter list. In

10
 Chapter 1 Java Essentials

listing 1.13, the call to overriddenMethod on line 16 goes to the method declared on line
nine, not to the method declared on line two.

1 public class Superclass {

2 public void overriddenMethod() {
3 System.out.println("Printed from overriddenMethod" +
4 " in superclass");
5 }
6 }
7
8 public class Subclass extends Superclass {
9 @Override
10 public void overriddenMethod() {
11 System.out.println("Printed from overriddenMethod" +
12 " in subclass");
13 }
14
15 public static void main(String[] args) {
16 Subclass subclass = new Subclass();
17 subclass.overriddenMethod();
18 }
19 }

Listing 1.13 overriddenMethod in Superclass is overridden by the method with the

same name in Subclass. The printout of this program is Printed from overriddenMethod in
subclass

Do not confuse overriding with overloading, which is to have methods with same name but
different signatures, due to different parameter lists. This has nothing to do with inheritance.
The keyword super always holds a reference to the superclass. It can be used to call the
superclass from the subclass, as illustrated on line ten in listing 1.14.

1 public class Superclass {

2 public void overridenMethod() {
3 System.out.println("Printed from Superclass");
4 }
5 }
6
7 public class Subclass extends Superclass {
8 public void overridenMethod() {
9 System.out.println("Printed from Subclass");
10 super.overridenMethod();
11 }
12
13 public static void main(String[] args) {
14 Subclass subclass = new Subclass();
15 subclass.overridenMethod();

11
 Chapter 1 Java Essentials

16 }
17 }

Listing 1.14 Calling the superclass from the subclass. This program prints Printed
from Subclass, followed by Printed from Superclass

To declare a class means to define a new type, therefore, the class named Subclass of
course has the type Subclass. When inheriting, the subclass will contain all methods and
fields of the superclass. Thus, the subclass will also have the type of the superclass, the
subclass in fact becomes also the superclass. This means that an instance of the subclass can
be assigned to a variable of the superclass’ type, see line 18 in listing 1.15. When a method is
called, as on line 19, the assigned instance is executed, not the declared type. This means the
method call goes to the method declared on line ten, not to the method declared on line two.

1 public class Superclass {

2 public void overriddenMethod() {
3 System.out.println("Printed from overriddenMethod" +
4 " in superclass");
5 }
6 }
7
8 public class Subclass extends Superclass {
9 @Override
10 public void overriddenMethod() {
11 System.out.println("Printed from overriddenMethod" +
12 " in subclass");
13 }
14
15 public static void main(String[] args) {
16 Subclass subclass = new Subclass();
17 subclass.overriddenMethod();
18 Superclass superclass = new Subclass();
19 superclass.overriddenMethod();
20 }
21 }

Listing 1.15 Calling a method in an instance of the subclass, that is stored in a field of the
superclass’ type. This program prints Printed from overriddenMethod in subclass, followed by
Printed from overriddenMethod in subclass

12
Chapter 2

Introduction
Before starting with object oriented analysis and design, it is necessary to understand how
those activities fit in the software development process. This chapter gives a general under-
standing of different activities performed in a programming project, and explains when and
why to do analysis and design.

2.1 Why Bother About Object Oriented Design?

Being able to change the front door of my house does not make me a carpenter, being able
to change spark plugs of my car does not make me a car-mechanic. Similarly, being able to
write a program that works when run by myself, on my own computer, does not have much
to do with being a professional software developer. On the contrary, professional software
development means to write code that can be maintained, changed and extended, in order to
meet the user’s expectations for a long period of time. This should hold even if developers
working with the code leave, and new developers arrive. To write such code, we need the
principles of object oriented design.
To be more specific, the goal of object oriented design is to write code that enables changing
the application’s behavior by changing as little code as possible, and absolutely only code
performing the task that shall be changed. It shall also be possible to extend the application’s
functionality without having to modify existing code. To reach this goal, the code must have
two important properties. First, it must be flexible, which means changes in one part of the
code does not require further changes in other parts of the program. Second, it must be easily
understood, structure and function shall be evident to anyone who reads the code.

2.2 Software Development Methodologies

Many software development projects have faced serious problems, for example being too ex-
pensive, being delayed or producing bad software due to bugs or lack of functionality. To
remedy these problems, there are many different sets of guidelines describing how to organize
a programming project the best way. Such a set of guidelines is called a software development
methodology. This section covers some important principles agreed on by all commonly used
software development methodologies.

13
 Chapter 2 Introduction

Software development must be iterative. During an iteration a limited amount of new

functionality is developed, or existing functionality is modified, or bugs in the code are cor-
rected, or some combination of these. What is important is that the work is completely finished
when the iteration is over. Iterations shall be relatively short, typically one or two weeks. The
reason for working like this is that it is only at the end of an iteration we really know the status
of the program being developed. There is no point in claiming that something is almost done,
either it is done, 100 percent ready, or it is not done. Each iteration is like a mini project,
which contains modifying requirements on the program, analysis, design, coding, testing, in-
tegrating new code with previously developed code, and evaluating the result together with
clients and/or users.
Manage risks early in the project. Code that is difficult to develop must be developed
during the first iterations. If not, it will be very difficult to make a reliable time plan for the
rest of the project, since we postpone work we do not really know how to perform. It might
even be impossible to write the difficult code. In that case, all work done in the project before
this is discovered is wasted, since either the project must be canceled or the program must be
rewritten.
Be prepared for changes. It is not possible to write a perfect specification of the program
before development starts. Both clients and the developers will come up with new, or changed,
ideas when they see the program. Therefore, there must be a procedure for managing changing
requirements. Actually, changes should be encouraged by working close to the client and
regularly demonstrating and discussing the program. This way the final result will be much
better and clients much happier than if developers try to oppose changes and force clients to
make up their minds once and for all.
Write extensive tests, and run them often. To make it easy and quick to run tests, they
should be automated. That means there should be a test program which gives input to the
program under test, and also evaluates the output. If a test passes, the test program does not
do anything. If a test fails, it prints an informative message about the failure. With extensive
tests that cover all, or most, possible execution paths through the program with all, or most,
possible variable values, it is guaranteed that the program works if all tests pass. This is a very
good situation, one command starts the test, which tells if the program under test works or, if
not, exactly which problems there are. This makes it easy to change the program, the test will
immediately tell if it still works after the change. Without tests, on the other hand, it will be a
nightmare to change the code since there is no certain way to tell if it still works.

2.3 Activities During an Iteration

Independent of software development methodology being used, the following activities are
typically performed during each iteration.
Requirements analyses is the process of identifying required functionality of the software
being developed. This process can not be finished early on in the project, but must be contin-
ued in each iteration. This is because clients can not be expected to know in detail what the
program shall do, before development starts. Both users and developers will come up with
new, or changed, ideas when trying early versions of the program. Therefore, it is import to

14
 Chapter 2 Introduction

work close to users and frequently discuss the functionality. In particular, each iteration shall
start with discussing requirements. Requirements analyses is not covered further in this text.
Analysis means to create a model, a simplified view, of the reality in which the system
under development shall operate. That model will consist of classes, attributes, etc. However,
it shall not describe the program that shall be developed, but rather describe the reality in
which the program shall operate. The purpose is to gain a better understanding of this reality,
before thinking about the program. Analysis is covered in chapter 4.
Design is an activity where we reflect on the code that shall be developed and create a plan
that gives a clear understanding of which classes and methods the code will contain, and how
they will communicate. To write a program without a plan is as inadequate as building a house
without a plan. Design is introduced in chapter 5.
Coding is of course the most important part of development, it is code quality alone that
decides if the program works a intended. The other activities have no other purpose than to
improve the quality of the code. This does not mean that the other activities can be neglected, it
is impossible to develop code of high quality without carefully performing all other activities.
Guidelines for writing high-quality code are covered in chapter 6.
Testing shall, as described above, be automated and extensive. Tests that are easy to execute
and clearly tell the state of the program are extremely valuable. They facilitate development
immensely since they make developers confident that the program works, also when changing
or adding code. Testing is covered in chapter 7.
Integrate means to add newly developed code to a repository with all previously devel-
oped code, and to verify that both new and previously developed code still work as intended.
The bigger the program and the more developers involved, the harder this process is and the
more important that it is well defined how to do it. Extensive and automated tests help a lot.
Integration is not covered further in this text.
Evaluation of code that was written during an iteration, is an important last activity of the
iteration. An iteration can not be ended without demonstrating the program to the clients and
gathering their opinions. The client’s opinions are added to the requirements and are managed
in coming iterations. This is not covered further in this text.
These are the main activities performed during each iteration, a typical iteration length is
one or two weeks. However, each developer shall also have a smaller, personal iteration,
which consists of designing, coding, testing and integrating. These four activities make an
indivisible unit of work, coding shall never be done alone without the other three activities.
Design is needed to organize the code and make sure it has the two required properties being
easy to modify and being easy to understand. Testing is needed to make sure the code works
as intended. Tests are also needed to show if the code still works as intended after coming
iterations. Integration with other code is needed because code parts are of no use unless they
work together.

2.4 Unified Modeling Language, UML

Both analysis and design result in plans. The results of analysis are plans of the reality modeled
by the program and the results of design are plans of the code that shall be written. These

15
 Chapter 2 Introduction

plans must contain symbols of classes, methods, etc, and to understand each other’s plans we
must agree on the symbols being used. To define those symbols is the purpose of the unified
modeling language, UML. UML is a vast standard, this text covers only the small fraction
needed to draw the plans that will be developed here.
UML defines different types of diagrams and the symbols that can be used in each of those
diagrams. Here, we will use class diagrams to give a static picture of something, and sequence
or communication diagrams to illustrate events following each other in time. When using
UML, it is important to understand that it does not say anything about the meaning of the
diagrams or symbols. For example, during analysis we use classes in a class diagram to
illustrate things in the reality. During design we use the same class symbols in class diagrams
to illustrate classes in an object oriented program. Thus, a class symbol can represent an
abstraction in the reality, a class in an object oriented program, or any other thing we choose
to let it represent. UML just defines what the symbol looks like.

16
Chapter 3

The Case Study

This text uses a car rental as case study to illustrate concepts and activities. More specifically,
the implemented functionality is RentCar, which describes what happens when a customer
arrives at the car rental office to rent a car. The requirements specification follows below.

3.1 Basic Flow

The basic flow, also called main success scenario, describes a sequence of events that together
make up a successful execution of the desired functionality, see figure 3.1.

1. The customer arrives and asks to rent a car.

2. The customer describes the desired car.
3. The cashier registers the customer’s wishes.
4. The program tells that such a car is available.
5. The cashier describes the car to the customer.
6. The customer agrees to rent the described car.
7. The cashier asks the customer for name and address, and also for the driving license.
8. The cashier registers the customer’s name, address and driving license number.
9. The cashier books the car.
10. The program registers that the car is rented by the customer.
11. The customer pays, using cash.
12. The cashier registers the amount payed by the customer.
13. The program prints a receipt and tells how much change the customer shall have.
14. The program updates the balance.
15. The customer receives receipt, change and car keys.
16. The customer leaves.

Figure 3.1 The basic flow of the RentCar case study.

17
 Chapter 3 The Case Study

3.2 Alternative Flows

An alternative flow describes a deviation from the basic flow. This requirements specification
has currently only one alternative flow, figure 3.2, which describes what happens if there is no
car matching the customer’s wishes.

4a. The program tells that there is no such car available.

1. The cashier tells the customer that there is no matching car.
2. The customer specifies new wishes.
3. Execution continues from bullet three in basic flow.

Figure 3.2 An alternative flow for the RentCar case study.

18
 Part II

Course Content

19
Chapter 4

Analysis
The purpose of analysis is to create a model, a simplified view, of the reality in which the
system under development shall operate. That model will consist of classes, attributes, method
calls, etc. However, it shall not describe the program being developed, but rather the reality
in which that program operates. The purpose is to gain a better understanding of this reality
before thinking about the program. This chapter shows how to develop a domain model and a
system sequence diagram. It also covers the UML needed for those two diagrams.

4.1 UML
This section introduces the UML needed for the domain models and system sequence dia-
grams drawn in this chapter. The UML diagrams used are class diagram and sequence dia-
gram. More features of these diagrams are covered in following chapters.
It can not be stressed enough that UML does not say anything about the meaning of dia-
grams or symbols. For example, a UML class in a UML class diagram is just that: A UML
class. It can represent something in the real world, like a chair, it can represent something in
!
a program, like a Java class, or it can represent something completely different.
When drawing a UML diagram, the meaning of the diagram and its symbols must be de-
fined. That is why specific diagrams have specific names, for example domain model. When
a diagram is given a well-defined name, everyone knows what it depicts and what its symbols
represent.

Class Diagram
A class diagram gives a static picture of something. It shows no flow or progress in time,
but only what classes there are, what they contain and how they are connected to each other. !
There is no notion at all of time in a class diagram.
The content of a class diagram might be a snapshot showing how things look at a particular
instant in time, or it might be the sum of everything that has existed during a specific time
interval, or it might be everything that will ever exist. This must be defined by the diagram
author.

20
 Chapter 4 Analysis

A class in UML means the same thing

as a class in an object oriented program,
a concept or idea not associated with any
(a) (b) (c) specific instance of that concept. A class
name is always a noun in singular. A class
Figure 4.1 Symbols for a class in a class diagram. Person specifies what a person contains.
(a) Without attribute and operation compartments.
(b) Empty attribute and operation compartments. It does not say anything about specific in-
(c) With an attribute. stances, like the persons me or you. Fig-
ure 4.1 shows three possible ways to draw
a class in a class diagram. The first example, figure 4.1a, specifies only the name of the class,
MyClass. The second, figure 4.1b, also specifies only the name. Here, however, there are
two empty compartments below the name. The upper of these is for specifying attributes.
As in object oriented programming, an attribute defines a value that can be attached to an in-
stance of the class. The bottom compartment, which is empty in all classes in figure 4.1, is
for operations. During analysis, there will not be any operations, therefore this compartment
will always be empty in this chapter. Finally, figure 4.1c shows a class that has the attribute
myAttribute.
Classes can have associations with other classes.
An association between two classes means that in-
stances (objects) of those classes are linked. If the
classes depicts classes in an object oriented program,
(a)
it means that one object has a reference to the other
object. If the classes depicts entities in the real world,
it means that instances have some kind of relation.
(b) Figure 4.2 shows some ways an association can be
illustrated in a class diagram. Figure 4.2a shows an
association with a direction. When drawn like that,
with an arrow, the association exists only in the di-
(c) rection of the arrow, Flight has an association with
Passenger, but not vice versa. There can be arrows
on both ends, meaning that both classes have an as-
sociation with the other class. There can also be no
(d)
arrow at all, which means that direction is not consid-
Figure 4.2 Associations ered. If there is no arrow at all, the diagram author
(a) Unidirectional chose not to tell the direction of the association.
(b) Unidirectional named In figure 4.2b, the association has a name, to clar-
(c) With name direction
(d) With multiplicity ify its meaning. If there is a name, the sequence ori-
gin class name, association name, target class name
should make sense and convey a message illustrating
the interaction of those three elements. This means the association name shall be a verb. In
figure 4.2b, the message is Flight transports Passenger.
The black triangular arrow in figure 4.2c shows in which direction the class-association-
class sequence shall be read, it does not tell anything about the association’s direction. It is
up to the diagram author to decide if such black triangles shall be used or not. They are most

21
 Chapter 4 Analysis

commonly used if class-association-class shall be read from right to left, or bottom up.
UML has different arrows, with different meaning. The arrows must look exactly as in
figure 4.2.
!
Figure 4.2d tells how many instances of each class are involved in the association. In this
example, there is exactly one instance of Flight, and five to fifty instances of Passenger.
This means the passengers travel with the same flight, which can take a maximum of fifty
passengers. Also, the flight will not take place if there are less than five passengers. It is pos-
sible to use the wildcard, *, when specifying the number of instances. It means any number,
including zero.

Sequence Diagram
A sequence diagram shows how instances send messages to each other. The UML term is
message, not method call. The messages in the diagram form a sequence of events, that
happen in a specified order in time.

(a) (b) (c)

Figure 4.3 Instances and messages in sequence diagram

(a) The instance myObj of MyClass
(b) Messages with activation bar
(c) Messages without activation bar

Figure 4.3a shows how to draw an instance. The word before the colon, myObj, is the name
of the instance and the word after the colon, MyClass, is the name of the class. Both names
are optional. The dashed line, called lifeline, is where messages to and from the instance are
anchored.
Figure 4.3b shows communication between two objects. Time flows from top to bottom, the
first message is bookSeat, which is followed by checkInLuggage. The message bookSeat
has a return value, which has the name accepted. The message checkInLuggage has a pa-
rameter, which has the name wheight. The thicker parts of the lifelines are called activation
bar, and means the instance is active during that period in time. If the sequence diagram de-
picts an object oriented program, the extent of an activation bar corresponds to execution of
a method. Figure 4.3c illustrates exactly the same as 4.3b, but without activation bars. This
format is preferred if it is not important to show when instances are active.

22
 Chapter 4 Analysis

Remember that different arrows have different meaning. The arrows must look exactly as
in figure 4.3. Generally, most things in UML are optional, but if used they must look exactly !
as defined in the specification.

(a) (b)

(c)

Figure 4.4 Conditions and loops in sequence diagram

(a) An if statement (b) An if else statement (c) A loop

Flow control is illustrated with combined fragments, which are the boxes drawn around
the messages in figure 4.4. A combined fragment consists of an interaction operator and an
interaction operand. The operators used here are opt, which illustrates an if statement, see
figure 4.4a; alt, which illustrates an if else statement, see figure 4.4b; and loop, which
illustrates an iteration, see figure 4.4c. The operands are the boolean expressions in square
brackets. In this example, figure 4.4a says that the passenger checks in luggage if the operand
hasLuggage is true. Figure 4.4b says that the passenger checks in luggage if hasLuggage
is true, and checks in without luggage if hasLuggage is false. Finally, figure 4.4c says that
the steward continues to serve meals while unservedPassengers is true. UML does not
specify operand syntax, any text is allowed in an operand.
To avoid confusion, it is always important to follow naming conventions. UML has mul-
tiple sets of naming conventions, the conventions used here are the same as in Java. Class
names are written in pascal case, LongDisctanceFlight; object names, attribute names, !
method names and variable names are written in camel case, economyClassPassenger,
luggageCount, checkInLuggage, unservedPassengers.

23
 Chapter 4 Analysis

Notes
Both class and sequence diagrams (and all other UML dia-
grams) can have comments, see figure 4.5. A comment is an
explaining text, a note to the reader, that is not part of any of
the elements in the diagram. Comments are often called notes
in UML. A note is anchored to the element it explains with a
dashed line, as in figure 4.5.
Figure 4.5 A UML comment

4.2 Domain Model

The domain model, DM, is a model of the reality that shall be represented in the program under
development. A UML class diagram is used to construct the domain model. The elements in
the DM are not classes in an object oriented program, but instead things that exist in reality.
Therefore, it might be better to call them entities instead of classes. The DM is a very good
tool for discussions about the program that is being developed. It can ensure that all parties
(developers, clients, users, etc) share a common view of the tasks of the program. Although
the DM does not depict the program, it will still prove to be very useful when constructing the
program. Since it is a model of the reality being modeled in software, it is highly likely that
the program will, when ready, have many things in common with the domain model.

Step 1, Use Noun Identification to Find Class Candidates

The first, and most important, step when creating a domain model, is to find as many class can-
didates as possible. Two complementary methods are used to find classes, noun identification
and category list, use both methods.
It is far more common to have too few classes than to have too many. It is also far more
problematic to have too few classes, since it is much easier to cancel existing classes than to !
find new ones.
The first method for finding class candidates, noun identification, means simply to identify
all nouns in the requirements specification, they are all class candidates. Below, in figure 4.6,
is the requirements specification for the Rent Car case study with all nouns in bold.
Since all words in bold are possible classes, each of them is drawn as a class in the first draft
of the domain model, figure 4.7. Remember that class names shall always be in singular.

Step 2, Use a Category List to Find More Class Candidates

The second method to find class candidates is to use a category list. It is a table where each
row specifies a category a class may belong to. The purpose is to stimulate the fantasy, thereby
to help find classes that are not nouns in the requirements specification.
The purpose of the category list is not to sort classes. There is no point in entering classes
already found during noun identification. There is also no point in spending time thinking NO!
about which row is correct for a certain class candidate.

24
 Chapter 4 Analysis

1. The customer arrives and asks to rent a car.

2. The customer describes the desired car.
3. The cashier registers the customer’s wishes.
4. The program tells that such a car is available.
5. The cashier describes the car to the customer.
6. The customer agrees to rent the described car.
7. The cashier asks the customer for name and address, and also for the driving
license.
8. The cashier registers the customer’s name, address and driving license number.
9. The cashier books the car.
10. The program registers that the car is rented by the customer.
11. The customer pays, using cash.
12. The cashier registers the amount payed by the customer.
13. The program prints a receipt and tells how much change the customer shall have.
14. The program updates the balance.
15. The customer receives receipt, change and car keys.
16. The customer leaves.
4a. The program tells that there is no such car available.
1. The cashier tells the customer that there is no matching car.
2. The customer specifies new wishes.
3. Execution continues from bullet three in basic flow.

Figure 4.6 The RentCar scenario, with nouns in bold.

Figure 4.7 The first draft of the domain model, after noun identification

There are many different proposals for categories. Here, the following quite short and
simple set is used,

• Transactions, selling or buying a product or service

• Products or services, what is sold or bought in the transaction
• Roles of peoples and organizations involved in the transaction
• Places, maybe where a transaction is performed
• Records of a transaction, for example contract, receipt
• Events, often with a time and place
• Physical objects

25
 Chapter 4 Analysis

• Devices, are probably physical objects

• Descriptions of things
• Catalogs, where the descriptions are stored
• Systems, software or hardware that is collaborating with the system for which we are
creating the DM
• Quantities and units, for example length, meter, currency, fee
• Resources, for example time, information, work force

The best way to create a category list is to simply consider each row in the category list
and try to imagine class candidates belonging to that category. Write down all classes that are
found, at this stage it is not interesting if the class is already listed or if it is relevant. Table 4.1
is a category list for the Rent Car case study.

Category Class Candidates

Transactions Rental, Payment, Insurance
Products, Services Car, CarKey, Rental
Roles, People, Organizations RentalCompany, Customer, Cashier
OfficeAddress, CustomerAddress,
Places
CarPickupLocation, CarLeaveLocation
Records RentalAgreement, Receipt
Events Rental
Physical objects Car, CarKey, Office
Devices
Descriptions RentalCondition, CarDescription
Catalogs CarCatalog, RentalCatalog
Systems
DrivenDistance, Kilometer, FixedCost,
Quantities, units
KilometerCost, InsuranceCost, Amount, Currency
Resources

Table 4.1 Category list for the Rent Car case study.

Next, all class candidates are added to the domain model, which now looks like figure 4.8.

Step 3, Choose Which Class Candidates to Keep

A question that always tends to be raised is how much it is meaningful to add to the require-
ments specification. For example, the Rent Car specification does not say anything about
insurances, which seems to make it a bit far-fetched to include the classes Insurance and
InsuranceCost. In a real project, this should be discussed with the customer. It might be
that something is missing in the specification. Here, there is no customer, we have to decide
on our own. Remember that it is much better to have too many than too few classes in the DM.

26
 Chapter 4 Analysis

Figure 4.8 The domain model, with classes from the category list added

Also remember that it is impossible to create a perfect model, there is a limit to how much
time it is meaningful to spend. Therefore, if it is really unclear if a class shall be removed or
not, just let it stay, at least for now.
Now consider figure 4.8, is there something that ought to be changed, in order to make the
DM clearer?

• The class Address is no longer needed, since there are the classes OfficeAddress
and CustomerAddress, and no more addresses need to be specified.

• The class Program was created since the requirements specification stated what the
program under development should do, but should it really be included in the DM?
The argument against is that the DM shall show only the reality, if Program is kept,
we have started to think about programming. In fact, all the other classes are a model
that shall be present inside the program. The argument for, on the other hand, is that
the program is in fact present in the reality. If a person, completely ignorant regarding
programming, where to write down all entities present in the rental office, the list would
include program (or system or something similar), since the cashier obviously interacts
with the computer.
This problem has no definite answer, it can be discussed endlessly. However, since this
text is a first course in analysis, Program is removed. The unexperienced developer
easily falls into the trap of modeling the program, instead of the reality, if the class
Program is present.

That is enough for now. If there are more irrelevant classes, they can be removed later,
before the DM is finalized.

27
 Chapter 4 Analysis

Step 4, Decide Which Classes Fit Better as Attributes

An attribute is not an entity of its own, but instead a property of an entity. Some classes should
not remain classes, but instead be turned into attributes. A simple, but very useful, guideline is
that an attribute is a string, number, time or boolean value. A class that contain just one such
value is a strong candidate to become an attribute. Another important rule is that an attribute
can not have an attribute. Consider for example a class Address. It can be represented as
a string, and is therefore a candidate to become an attribute. On the other hand, it might be
convenient to split it into street, zip code and city. If that is preferred, Address must remain
a class, to be able to contain the attributes street, zipCode and city. A third rule is that
when it is hard to decide if something is an attribute or a class, let it remain a class. Better to
have too many classes than too few.
Now consider the domain model of figure 4.8 (remember Address and Program where
removed). Which classes fit better as attributes?

• DrivingLicenseNumber is a number (or a string), it can become an attribute of

DrivingLicense.

• Name is a string. Unless it is relevant to split it into first name and last name, it can be
an attribute of Customer. It can also be an attribute of Cashier, if needed.

• Should CarKey be an attribute of Car? It is true that CarKey can be considered to be

strongly associated with Car, but it is not obvious that CarKey is a string, number, time
or boolean. Therefore, it remains a class.

• Amount is a number, and could become an attribute of Cash and Balance. But then
what about Currency? Is that not a string that should be an attribute of Amount? This
is something that should be discussed with the customer, but now let’s just decide we
do not need to keep track of currencies. Therefore, Currency is removed and Amount
becomes attributes of Cash and Balance, and also of Change, Payment, FixedCost,
KilometerCost and InsuranceCost.

• OfficeAddress and CustomerAddress could be attributes of Office and Customer,

but according to the reasoning above about addresses, we keep them as classes. These
classes should be associated with Office and Customer, respectively. That will be
done below, when considering associations. However, there is no point in creating a
new class for each new address. Instead, OfficeAddress and CustomerAddress are
removed, and the single class Address is reintroduced.

• Is CarDescription an attribute of Car? No, since it most likely contains quite a lot
of information, like model, model year, size, etc. All this can not be represented as a
single string.

• Kilometer is is the unit of the quantity DrivenDistance. Provided there are no

other units, it can be removed. DrivenDistance is a number, it becomes an attribute
of Rental.

28
 Chapter 4 Analysis

• InsuranceCost is a number, it becomes an attribute of Insurance. KilometerCost

and FixedCost are also numbers, they are turned into attributes of RentalCondition.

Figure 4.9 The domain model with attributes

That is enough, remember that there is no point in minimizing the number of classes. The
domain model with attributes is depicted in figure 4.9.

Step 5, Add Associations

The purpose of associations in the domain model is only to clarify. Therefore, add only associ-
ations that actually do clarify the DM. It is almost always possible to find more, no matter how
many there are already. At some point, where more associations are just confusing, it is simply
necessary to stop. Try always to name the association, without name it hardly clarifies at all.
Try to avoid the names has and hasA since it quite obvious that a class with an association to
another class has an instance of that class. More or less all associations could be named has.
It is strictly forbidden to create names that must be read as parts of a chain of associations, for
example Passenger checksIn Luggage at Counter, which is class-association-class-
association-class. It is impossible for the reader to now where such sentences start and end, the
reader would probably try to read just Luggage at Counter, which does not make sense.
The sequence must always be exactly class-association-class, and the association name must
start with a verb.
Multiplicity is often just confusing, add multiplicity only if it clarifies the DM. Do not
specify direction, trying to understand the direction of an association in the DM often leads
to long and meaningless discussions. Also, if two entities in the reality are associated, it is
almost always bidirectionally. Finally, there should be at least one association to each class.
If it is hard to find an association to a certain class, or if there are different sets of internally
associated classes that are not joined by associations, it is a sign that there is something wrong
with the DM.

29
 Chapter 4 Analysis

Start with the most central associations. Since it is all about renting a car, that could be for
example Customer performs Rental, Car isRentedIn Rental, Payment pays Rental
and Car isOwnedBy RentalCompany. Then continue, following the guidelines above. The
result can be seen in figure 4.10.

Figure 4.10 The domain model with associations

Insurance, CarPickupLocation and CarLeaveLocation were removed since they

were not mentioned in the requirements specification, and the DM is becoming quite big and
messy. Also Cash was removed. Whether to include it or not is a question of how detailed
a payment record shall be. Is it of interest to know how much cash the customer gave to the
cashier?
The class Address has no association. This is OK for classes that exist just to group data,
and do not have a specific meaning but are used in many places. Examples of such classes are
Address, Name, Amount and Coordinate. The reason is that the DM would be unclear if
associations where added to all classes using such data containers. Instead, usage is illustrated
by adding the data containers as attributes to classes using them. In figure 4.10, for example
Office and Customer has an attribute address, showing that they use the Address class.

30
 Chapter 4 Analysis

Step 6, Anything To Change?

To create the domain model is an iterative process. New classes might be found while con-
sidering attributes and associations, attributes might be changed while adding associations, etc.
This case study was also performed iteratively, for example was the need for the RentalCatalog
class discovered when adding the association between CarDescription and CarCatalog.
Therefore, it is good practice to reconsider the entire DM when done with associations. Here,
there is no obvious need for changes, the DM of figure 4.10 becomes the final version.

Common Mistakes
Since creating a domain model is a matter of discussion and, at least to some extent, a matter
of opinion, it might be difficult to assess the quality of the resulting DM. There are many
ways to create a good model, but also many ways to create a bad model. This section explains
some typical mistakes, resulting in a model of low quality. Such a model might not be plainly
wrong, but is of little help to the developers.
The first, and most obvious, mistake is not to model reality, but instead regard the DM
as a model of a program. This normally also means that some notion of time is assigned
to the DM. Things are thought happen in a sequential order, whereas a DM (or any UML
class diagram) says absolutely nothing about time or order of events. A class Program or
System often becomes essential in such a “programmatic DM”, but be aware that the role NO!
of the program can be assigned to any other class as well. Also, an association is considered
to be some kind of method call, instead of a relation. Finally, the actor, which is the cashier
in the case study, becomes the user of the program. Figure 4.11 shows an example of a
“programmatic domain model” where the class Office represents the program.
Another, less obvious, mistake, is to create a DM that correctly models the reality, but does
not convey any information besides what is already in the requirements specification. In such
a “naïve domain model”, the actors, customer and cashier in the case study, become central
classes with many outgoing associations. Other classes tend to be associated only with one
of the actors. This kind of DM is in fact just a visual representation of the specification. NO!
It focuses on what the actors do, modeling flow, instead of giving a static picture of what
exists. This might not be completely wrong, but adds little value to what already exists, in
text. Figure 4.12 is an example of a naïve DM, compared to the DM of figure 4.10, it does
not say much.

31
 Chapter 4 Analysis

NO!

Figure 4.11 This is not a correct domain model. The modeler has tried to create a program, instead of modeling
the reality in which the program acts.

NO!

Figure 4.12 This domain model does not add any extra value, or new information.

32
 Chapter 4 Analysis

NO!

Figure 4.13 This extract of the RentCar case study has a class, Rental, with unnecessarily many associa-
tions.

It is best to avoid creating a “spider-in-the-web” class. Such a class has associations to

many other classes, while other classes have few associations, especially to classes besides
the spider class. A DM with a spider-in-the-web class may still be valuable, but would
probably be of higher value if associations were more evenly distributed. The central class,
with many associations, is often difficult to understand, since it seems to have many different
roles. Also the roles of the peripheral classes, with very few association, might be difficult
to understand. They seem to be just “data containers”, like primitive values, without any
real role to play. It is very difficult to give a definite rule for when a class has become a
NO!
spider class. A coarse guideline could be that a class should not have more than four or five
associations, but that depends on the size and layout of the entire DM. A spider class is made
less central by moving an association from it to another class, which is in turned associated
with the spider class. As an example, consider the Rental class in figure 4.10. It has four
associations, but had five in a previous version of the DM, where it looked as in figure 4.13.
The association with RentalCondition was moved to RentalAgreement.
Remember that an attribute is a property of a class, not a part. It is a mistake to model parts
as attributes. As an example, consider a house, which has windows, doors and rooms. It is
not correct to model those as attributes of the house, which is illustrated by the fact that it is
very hard to regard a window, door or room as a string, number, time or boolean, which was
the guideline for creating an attribute. This reasoning is illustrated in figure 4.14. It is not NO!
always obvious whether something is a property or a part. For example, why is a window a
part of a house, while a name is a property of a person? A rough guideline can be that the
relationship is part when the whole actually is constructed of the part, otherwise property is
better.

33
 Chapter 4 Analysis

(a) (b) NO!

Figure 4.14 In (a), window is, erroneously, modeled as an attribute of house. It is better to model it as a class,
as in (b).

4.3 System Sequence Diagram

The system sequence diagram, SSD, is a sequence diagram that shows the interaction between
the system under development and the actors using it. An actor is any person or other system
giving input to or receiving output from the system. The SSD must not show anything about
the system’s internal structure, the entire system must be modeled as one single object. Apart
from this System object, there is one object for each type of actor. The messages, that is oper-
ation that actors can perform on the system, are called system operations. A well constructed
SSD simplifies development a lot, since it shows exactly what the system can be told to do,
and what response it shall give. Strictly speaking, creating an SSD is not part of the analysis,
but instead belongs to gathering requirements. Here, we consider it under the analysis section
since it is a preparation for program construction.
Do not confuse system sequence diagram with sequence diagram. Although the names
are similar and an SSD is created using a sequence diagram, they are far from synonyms.
Sequence diagram is the UML name of a kind of diagram used to illustrate how objects
exchange messages, as explained above. It can be used to illustrate any kind of interaction.
!
One specific way to use a sequence diagram is to create an SSD, which is a diagram that
illustrates how actors interact with a program, and nothing else.
While the domain model is very much a matter of discussion, the SSD is more straight
forward to create. It shall reflect the interactions of the requirements specification, no less and
no more. It is common to find errors or ambiguities in the specification when constructing the
SSD. In that case it might have to be revised, but it is not allowed to let the SSD deviate from
the specification.
Since the SSD shall show the interaction between the system and its actors, therefore the
first step is to define where the system ends, and which the actors are. In the case study, it is
obvious that we are not developing the cashier, but we are developing the thing the cashier in-
teracts with. Thus, the cashier becomes the actor. Then, look at the requirements specification
and identify what the cashier can tell the system to do, and how it responds. The specification
is repeated here, in figure 4.15, for the sake of convenience.
Bullets one and two do not contain any interaction between the actor (cashier) and the
system. Remember that it is completely uninteresting for the SSD what happens “outside” the
actor. Therefore, it would be wrong to include the customer.
In bullet three, there is an interaction that shall be included. A system operation shall
have a name that starts with a verb, and describes what is done. The system operation in
bullet three can be named searchMatchingCar. The name shall not describe what happens

34
 Chapter 4 Analysis

1. The customer arrives and asks to rent a car.

2. The customer describes the desired car.
3. The cashier registers the customer’s wishes.
4. The program tells that such a car is available.
5. The cashier describes the car to the customer.
6. The customer agrees to rent the described car.
7. The cashier asks the customer for name and address, and also for the driving license.
8. The cashier registers the customer’s name, address and driving license number.
9. The cashier books the car.
10. The program registers that the car is rented by the customer.
11. The customer pays, using cash.
12. The cashier registers the amount payed by the customer.
13. The program prints a receipt and tells how much change the customer shall have.
14. The program updates the balance.
15. The customer receives receipt, change and car keys.
16. The customer leaves.
4a. The program tells that there is no such car available.
1. The cashier tells the customer that there is no matching car.
2. The customer specifies new wishes.
3. Execution continues from bullet three in basic flow.

Figure 4.15 The requirements specification for the RentCar case study.

internally, in the system, searchInDatabase is therefore not an adequate name. This system
operation also takes parameters, namely the customer’s description of the desired car. We
could write a long list with these parameters, e.g., size, price, model, desired features (for
example air condition), etc. The downside of such a solution is that a lot of time would be spent
deciding exactly which parameters to include. Also, if the set of parameters would change,
we would have to change this system operation. And the set of parameters likely will change,
as development continues and the needs the system shall meet become clearer. Therefore,
it is better to use an object as parameter. This object, which can be called wishedCar, has
attributes that define the set of possible wishes. Note that types, whether primitive or classes,
are not of interest in the SSD. Now, exactly which the wishes are is not necessary to decide.
Also, if the set of possible wishes changes, that is an internal matter for the class of this object,
no changes are required to the system operation.
Now the system operation and its parameters are identified. Next, it must be decided if the
operation has a return value, and, if so, which return value? The answer is found in bullet four,
which tells that the system gives a positive answer to the search, and in bullet five, which tells
that the cashier describes the found car. Considering only bullet four, it might seem adequate
to use a boolean return value, but bullet five clearly states that the return value must include a
description of the found car. Therefore, also the return value can be an object, which can be
called foundCar. The system sequence diagram with this first system operation is depicted

35
 Chapter 4 Analysis

in figure 4.16. Note that activation bars are omitted, which is practice in system sequence
diagrams. It is not relevant to know when actors and systems are active. Also, trying to decide
this tends to lead to long and quite meaningless discussions.

Figure 4.16 The system sequence diagram, after the first system operation has been created.

Continuing the same way, bullet six in the requirement specification does not involve any
interaction with the system, neither does bullet seven. Bullet eight defines the second sys-
tem operation, which can be called registerCustomer. An alternative name could be
registerCustomerData, but it is usually unnecessary to include the word data, since more
or less all operations include data in some way. The parameters of this operation are name,
address and driving license number. These could very well be joined in an object, customer,
according to the reasoning above, for the searchMatchingCar system operation. But since
this parameter list is defined exactly in the specification, it is less likely that it changes in the
future. The latter alternative is chosen, since that clearly shows that the parameters are known.
Bullet nine is a system operation, bookCar. It is a bit unclear if it shall take any parameters.
It could be argued that it does not take any parameters, in which case the system must keep
track of the car returned in the searchMatchingCar operation, and book that same car. It can
also be argued that the car to book shall be specified in the bookCar operation. If so, the name
of the parameter shall be the same as the name of the return value of searchMatchingCar,
to show that it is in fact the same car. The latter alternative is chosen, mostly since the for-
mer would require us to remember that the car returned from searchMatchingCar must be
stored, and is therefore a bit unclear.
Bullet ten is no system operation. It describes work done internally, inside the system, and
does not involve any interaction with an actor. Bullet eleven also is no system operation, since
it takes place only between customer and cashier. Bullet twelve is a system operation, it can
be called simply pay, and take the parameter amount. The SSD now looks as in figure 4.17.
In bullet 13, a receipt is printed as a result of the pay operation. Does this mean receipt
is a return value of pay? The answer depends on if the printer is considered to be part of the
system under development (SUT), or not. If the printer is part of the system, the printing is an
internal matter and shall not be included in the SSD. The receipt then becomes a return value.
On the other hand, if the printer is not part of the SUT, it becomes an external system, called
from the SUT. This means there is an interaction between the SUT and an external entity,
which shall be included in the receipt. The latter alternative is chosen, mainly to illustrate how
such an interaction looks in the SSD, se figure 4.18.
Continuing, bullets 14-16 are either internal or external, and to not bring any interaction

36
 Chapter 4 Analysis

Figure 4.17 The system sequence diagram, with more system operations added.

Figure 4.18 The system sequence diagram, with call to external system.

between the SUT and its actors. Therefore, they do not generate any system operation. That
concludes the basic flow, next the alternative flow is considered. The alternative flow specifies
a loop, including bullets two, three and four in the basic flow. The iteration around these
bullets continues until a matching car is found. Note that the specification is incomplete, it
does not allow the customer to give up and leave without renting a car. Of course this must be
changed in coming iterations of the development. The loop can be modeled as in figure 4.19.
There are a two things worth highlighting regarding the loop. First, the guard noMatchingCar
is a free text boolean condition, it does not correspond to a boolean expression in the program.
The condition can become true because of an action taken by the system, the actor or some-
thing completely independent of both system and actor. Second, drawing the return value
foundCar inside the loop, as in figure 4.19, implies it can indicate both that a matching car
was found and that such a car was not found. How this is done is not shown is the SSD.

37
 Chapter 4 Analysis

Figure 4.19 The system sequence diagram, with a loop reflecting the alternative flow.

Common Mistakes
As mentioned above, there is not so much freedom in drawing the system sequence diagram
as there is in drawing the domain model. Many mistakes do not lead to a correct diagram of
less value, but instead to one that is plainly wrong. Before leaving the SSD, it is therefore wise
to make sure that none of the following common mistakes are made.
• Wrong kind of arrow.
• System operation, return value or parameter is missing.
• Operation name does not start with a verb.
• Operation name describes the system’s internal design, for example
searchInDatabase instead of search.
• Entities outside the actor are included. A typical version of this mistake would be to
include an object :Customer in the case study’s SSD.
• The object :System is split into more objects, showing the system’s internal design. NO!
As an example, it would be wrong to include an object :Car, :Rental or :Balance
in the case study.
• Loops or if-statements are not correctly modeled.
• External systems, like :Printer in the case study, are missing.
• To draw activation bars is not wrong, but it is discouraged since it tends to confuse,
rather than clarify.

38
Chapter 5

Design
The purpose of design is to reflect on the code that shall be developed and create a plan that
gives a clear understanding of which classes and methods the code will contain, and how they
will communicate. To write a program without a plan is as inadequate as building a house
without a plan. The created plan, that is the design, shall guarantee that the program becomes
flexible and easy to understand. Flexible means that it shall be possible to add new func-
tionality without having to change existing code, and to change existing functionality without
having to change any code besides that handling the actual functionality being changed. Easy
to understand means that developers not involved in originally creating the program, shall
be able to understand and maintain it, without rewriting anything or destroying the program
structure.
The result of the design is a plan in the form of UML diagrams, illustrating the details of
the program. Before those can be created, this chapter covers the necessary UML. After that,
three concepts are covered, that are necessary requirements for a design that is flexible and
easy to understand. Next comes an introduction to architecture, or, more specifically, how to
organize the program in subsystems. The last thing before doing the design of the RentCar
case study, is to present a step-by-step method for design.
It is not possible to create a design of a program without understanding how the design
can be implemented in code. Make sure you fully understand sections 1.2 and 1.3 before !
reading this chapter.

5.1 UML
This section introduces the UML needed for the design diagrams. Two new diagram types are
introduced, package diagram and communication diagram. Also, more features of class and
sequence diagrams, which where introduced in chapter 4.1, are covered.

39
 Chapter 5 Design

Class Diagram
To create a design class diagram, some features are
needed that have not been used previously in this text,
namely methods, visibility and types. Methods are de-
clared in the lowest compartment of the class symbol.
A method parameter’s type is written after the param-
(a)
eter, separated from the parameter by a colon. The
methods return type is written the same way, but af-
ter the entire method, see figure 5.1a. Static methods
(and attributes) are underlined, see figure 5.1b. The
visibility of a class member (method, attribute or any-
(b)
thing else defined in the class) defines an object’s level
of access to that member. For now, only two kinds of
visibility are considered, public and private. Any code
has access to a member with public visibility, while
only code in the declaring class has access to a mem-
ber with private visibility. The symbols + and - are
(c) used in uml to indicate public and private visibility,
respectively, see figure 5.1c.
Figure 5.1 Class diagram, illustrating:
(a) method (b) static members
(c) public and private visibility Package Diagram
The UML symbol package means just a grouping of
something. In a class diagram of a Java program, the
package symbol can mean a Java package. It can also
be used to illustrate a larger grouping, like a subsystem
consisting of many Java packages. Figure 5.2 shows
an example of a package diagram. The dashed line
means that something in somePackage is dependent Figure 5.2 A package diagram
on something in someOtherPackage. The diagram
does say anything about the extent or type of this dependency.

Sequence Diagram
This section explains some previously not covered features, needed to create design sequence
diagrams. Figure 5.3 illustrates some of these. First, the call to firstMethod is a found
message, which is specific in the sense that the caller is unspecified. This is normally used
when the origin of the message is outside the scope of the diagram, and shall not be described
in detail. The scope of the diagram is to describe what happens as a consequence of the call to
firstMethod, not to describe when or why that call is made.
Second, types are depicted as in a class diagram, following the parameter or method, sepa-
rated by a colon.

40
 Chapter 5 Design

Figure 5.3 Found message, types, activation bars, message to caller, constructor

Third, note the use of activation bars, which show the duration of a method. The bar begins
on the first line of a method and ends when returning from the method. There can be any num-
ber of activation bars overlapped simultaneously on the same object, since execution might be
inside any number of methods in the same object. For example, firstMethod in the object
someObj in figure 5.3 calls aMethod in otherObj, which in its turn calls someMethod in
someObj. Since at this point firstMethod has not yet returned, execution is inside both
firstMethod and someMethod, illustrated by the double activation bar of someObj.
Fourth, the call to methodInSelf illustrates a call where the caller and callee are the
same object. Also in this case there is a double activation bar, since execution is inside both
firstMethod and methodInSelf.
Last, a constructor call is illustrated with the
creation of newObj. This method must be called
ThirdClass, since a constructor always has the
same name as the class in which it is located.
Also, the return type must be ThirdClass, since
the newly created object of that type is returned
by the constructor. The text «create» above the
constructor call is a stereotype, which tells that
the element with the stereotype belong to a cer-
tain category of such elements. Here, it says that Figure 5.4 Static method and interaction use
the method ThirdClass belongs to the create
category, which means it is a constructor. A stereotype can contain any text, the diagram author
is free to invent new stereotypes. However, there are conventions, for example constructors
have, by convention, the stereotype «create» . It would seem that «constructor» would
be a more logical stereotype, but unfortunately «create» is used instead.
Figure 5.4 illustrates two more features of a sequence diagram. First, met2 is a static
method, which is illustrated with the stereotype «static» . Second, the box labeled ref is
an example of an interaction use. It tells there are more method calls where the box i placed,
those can be seen in a sequence diagram named SomeTask. A diagram should be split like
this when it becomes difficult to understand, or fit in a page, because of its size.

41
 Chapter 5 Design

Communication Diagram
A communication diagram serves exactly the same purpose as a sequence diagram, to il-
lustrate a flow of messages between objects. Both these types of diagrams are interaction
diagrams. Which type of interaction diagram to use is completely up to the creator, ev-
erything relevant for design can be illustrated in both types. The advantage of a sequence
diagram is that time is clear, since there is a time
axis (downwards), whereas a communication dia-
gram does not have a time axis but illustrates mes-
sage order by numbering the messages. The ad-
vantage of a communication diagram is that ob-
jects can be added both horizontally and verti-
cally, whereas a sequence diagram has all objects
beside each other and therefore tend to become
very wide.
Figure 5.5 shows a communication diagram. A
caller and callee are connected by a line, called
link. A message (method call) is illustrated by an
arrow along the link and the name of the message
(method), its parameters, types and return value.
The first call, metA, has index 1. If a called is
made from the method metA, it has number 1.1,
as illustrated by metB in figure 5.5. A second call Figure 5.5 A communication diagram
from metA has index 1.2 and so on. The order of
execution in figure 5.5 is thus 1, 1.1, 1.2, 1.2.1,
2, 3, 3.1. If there are more than one messages
between the same pair of objects, they are still connected by only one link, see calls 2, 3 and
3.1. Message 1.2 illustrates a constructor call and message 1.2.1 shows a message where caller
and callee are the same object.
Figure 5.6 illustrates conditional calls in mes-
sage 1 and 2, and iteration in message 2.1. A con-
dition is called guard and is specified in square
brackets. UML does not specify guard syntax,
any text or program statement is allowed. The as-
terisk in message 2.1 indicates iteration. It shall
be placed before the square bracket, *[i=1..n],
but that is unfortunately not possible in astah.
Therefore, it is included in the guard statement,
which is not correct UML.
Figure 5.6 Conditional call and iteration in com-
munication diagram

42
 Chapter 5 Design

5.2 Design Concepts

Much is written and said about software design in general and object-oriented design specifi-
cally, and there are many more or less complex solutions to various problems. Still, virtually
all design considerations and solutions are based on a few principles. The first and most fun-
damental of those, encapsulation, high cohesion and low coupling, are covered here.

Encapsulation
Encapsulation means that irrelevant internal details are hidden. In order to use a certain item,
for example a clock, it is not required to under-
stand exactly how it works internally, as in figure
5.7. Instead, it exposes an interface with every-
thing the user has to know. In case of the clock,
this is the current time.
To understand encapsulation in software, the
concept visibility must be clear. The visibility of
a declaration states where that declaration is visi-
ble. For now, it is enough to understand two kinds
of visibility, public and private. Public visibil-
ity, specified in Java with the modifier public,
makes the declaration visible to all parts of the
Figure 5.7 If you would have to understand all the program. Any piece of code, anywhere in the
internals of a clock to tell the time, it would entire program, can use a declaration with pub-
be very bad encapsulation. Image by FreeIm-
ages.com/Colin Adamson
lic visibility, no matter what is declared or where
it is declared. Private visibility, specified in Java
with the modifier private, means the declaration is visible only to code in the class in which
the declaration is placed.
1 public class TheClass {
Encapsulation relies on the difference
2 private int var;
between public interface and implemen-
3
tation. The public interface is code that 4 public TheClass(int var) {
is visible to all other code, that is, the 5 this.var = var;
declarations with public visibility. The 6 }
implementation consists of code not vis- 7
ible to all other code, that is, method 8 public void
bodies and declarations with private vis- 9 doSomething(String s) {
ibility. Something is part of the im- 10 anotherMethod(s);
plementation when access to that some- 11 }
thing can be controlled. It is possible to 12
tell exactly which code can access a cer- 13 private void
tain part of the implementation, it is not 14 anotherMethod(String s) {
required that no other code at all can ac- 15 //Some code
cess it. Also, there is no “neutral” code, 16 }
17 }
all code is either public interface or im-
Listing 5.1 Public interface in blue italic.

43
 Chapter 5 Design

plementation. As a first example, consider listing 5.1, where the public interface is marked
with blue italic print. The defining question is if this code is changed, can code anywhere in
the entire program be affected? If the answer is yes, it is part of the public interface. That is
why parameter and return value of the public method is marked in listing 5.1.
Continuing with slightly more complicated examples, consider listing 5.2. The static
modifier of PI is part of the public interface, since PI might be used in a statement like
MyClass.PI. If static is removed, that statement will not work. The status of the modifier
final is quite subtle. If a non-final field is made final, code might definitely break since it
will no longer be allowed to write to that field. If a final field becomes non-final, it is not
obvious that code will break. However, it would be very surprising if a (previously) final field
suddenly changed value. The conclusion of this reasoning is that it is safest to consider the
final modifier to be part of the public interface.
The type and name of PI are of course part of the public interface, but why not the value?
The answer lies in the promise of this field, which is specified in the comment. Whether the
value is 3.14, 3.1416 or has some other precision, it would still fulfill its contract, to be the
constant pi. If the comment had said “The constant pi with two decimals”, the value would
have been part of the public interface. Is this a bit of hairsplitting? Maybe, but then at least it
illustrates how one can reason when identifying the public interface.
The private constructor on line seven is not public interface, what matters is that it is pri-
vate. That it is a constructor is of no importance. Finally, the exception list on line eleven is
definitely part of the public interface. If it is changed, exception handling code might break.
This holds both for checked and unchecked exceptions.

1 public class MyClass {

2 /**
3 * The constant pi.
4 */
5 public static final double PI = 3.14;
6
7 private MyClass() {
8 //Some code.
9 }
10
11 public void aMethod() throws SomeException {
12 //Some code.
13 }
14 }

Listing 5.2 Illustration of public interface, which is in blue italic print.

44
 Chapter 5 Design

The point in making this distinction between public interface and implementation is that
the implementation can be changed anytime, without any risk of unwanted consequences.
Changing the public interface, on the other hand, is very dangerous since any code anywhere
might break. That might not be a big issue in a small program with only one developer.
However, in programs just slightly bigger, with more than one developer, changing the public
!
interface immediately becomes challenging. Those whose code break will not be very happy,
especially if it happens regularly or without notice. This is even more disastrous if the code
is part of a published API, where it is impossible to know who is using it.
As an example, consider the two methods in listing 5.3. They both have exactly the same
public interface, but implementations differ. It would be no problem at all to change between
the two implementations, code that calls multiplyWithTwo is completely independent of
whether the multiplication is done by straightforward multiplication or by shifting. The same
way, it is without any risk to change other parts of the implementation, for example name or
parameter types of a private method. It could be argued that it is not allowed to change the
implementation of multiplyWithTwo at free will, for example not to return operand
* 5. It is true that such a change can not be made, but that is because it changes the public
interface, since both name and comment become erroneous by such a change.

1 /**
2 * Doubles the operand and returns the result.
3 */
4 public int multiplyWithTwo(int operand) {
5 return operand * 2;
6 }
7
8 /**
9 * Doubles the operand and returns the result.
10 */
11 public int multiplyWithTwo(int operand) {
12 return operand << 1;
13 }

Listing 5.3 Two methods with the same public interface, but different im-
plementations.

In conclusion, it is essential that a public interface is well designed and as small as possible.
As soon as a program grows to any reasonable size, it becomes very difficult to change any
part of its public interface, to say the least. Many programs suffer from strange constructs
originating in public interfaces impossible to change.

45
 Chapter 5 Design

High Cohesion
Cohesion is a measurement of how well defined a
class’ knowledge and its tasks are, and how well
they fit together. The goal is that a class shall
represent one single abstraction, which is clearly
identified by the class name. Furthermore, the
class shall have knowledge about that abstraction,
not about anything else, and perform tasks related
to that abstraction, not anything else. When this
important goal is reached, the class has high co-
hesion.
Figure 5.9 shows two different designs of the
Figure 5.8 When there is high cohesion, the parts same program, one with low cohesion (figure
fit together and create a whole that is easy to
understand. Image by unknown creator [Public do- 5.9a) and one with high cohesion (figure 5.9b). In
main], via https://pixabay.com the low cohesion design, the Employee class has
the method getAllEmployees, which returns a
list of all employees. This means an Employee instance, which represents one single em-
ployee, knows all employees in the department. That is not relevant knowledge, instead, that
information fits better in a Department class, which reasonably shall know all employees
working at the department. This latter design, with higher cohesion, is illustrated in figure
5.9b. Also, the Employee in figure 5.9a has a method changeSalaryOfEmployee, which
can change the salary of any employee, not just that particular instance. The better design,
in figure 5.9b, has another version of this method, which means an instance can change only
its own salary. In conclusion, in the better design, an Employee instance knows about, and
performs operations on, only itself, while in the worse design, an instance knows about, and
performs operations on, any instance.

(a)

(b)

Figure 5.9 Two different designs of the same program:

(a) with low cohesion (b) with high cohesion

46
 Chapter 5 Design

Another example is given in figure 5.10, which illustrates a Car class. In the design with
lower cohesion, figure 5.10a, Car has methods and attributes which are more related to the
abstractions radio and engine. In the design with higher cohesion, figure 5.10b, those attributes
and methods are moved to the new, appropriately named, classes Radio and Engine. Of
course, as is the case with all designs, it can be argued that there are problems also with the
designs in figures 5.9b and 5.10b. Still, those two definitely have higher cohesion than their
low cohesion counterparts in figures 5.9a and 5.10a.

(a)

(b)

Figure 5.10 Two different designs of the same program:

(a) with low cohesion (b) with high cohesion

It is absolutely mandatory to always strive for high cohesion. If not, the program will be
difficult to understand, and also difficult to change, since code that is not really related will
be mixed together in the same class. The programmer can not relax even if the program, at
some point in time, has high cohesion. As more code is added, the program will eventually
!
get low cohesion if classes are not split. Therefore, always be on the guard for the possibility
to improve the design by introducing new classes, with a clearer responsibility.
Finally, although only classes have been discussed in this section, exactly the same rea-
soning applies to programming constructs of all other granularities as well. Also subsystems,
packages, methods and even fields must be continuously scrutinized regarding cohesion.

47
 Chapter 5 Design

Low Coupling
Coupling defines to which extent a class depends
on other classes. What is interesting is primarily
on how many other classes it depends, the type
of dependency is not of great interest; whether
method call, parameter, return value or something
else does not matter much. Low coupling means
there are as few dependencies as possible in the
program. It is not possible to tell a maximum al-
lowed number, what matters is that there are no
dependencies which are not required.
The main reason to strive for low coupling
is that if a class (depender) depends on another
Figure 5.11 If the UML diagram looks like a
bowl of spaghetti, there is too high cou- class (dependee), there is a risk that the depender
pling. Image by Katrin Baustmann [Public domain], via must be changed as a consequence of a change in
https://pixabay.com
the dependee. If for example a method name is
changed, also all classes calling that method must
be changed. The problem is bigger the less control the developer has of the dependee. For
example, it is a quite small problem if the program is small and developed by only one person,
but much bigger in a large program where developers far away might change the dependee.
The size of the problem is also defined by the stability of the dependee. The more often a class
is changed, the bigger problem to depend on it. For example, it is completely safe to depend on
classes in the APIs in the JDK, in the java.* packages, since they change extremely seldom.
Figure 5.12 shows two different designs of the same program, one with high coupling
and one with low coupling. In the version with unnecessarily high coupling, figure 5.12a,
HighCouplingOrder has a reference to HighCouplingShippingAddress. This is not
required since Order can get ShippingAddress from Customer. Therefore, this reference
can be omitted, as illustrated in figure 5.12b.

(b)
(a)

Figure 5.12 Two different designs of the same program:

(a) with high coupling (b) with low coupling

Another example of unnecessarily high coupling is found in figure 5.13a, which depicts a
typical “spider in the web” design, with a “spider” class that has references to many other,
peripheral, classes. The peripheral classes in such a design tends to have none or very few ref-
erences to other classes. The problem here is that the spider class normally becomes involved

48
 Chapter 5 Design

in all operations, thereby getting messy code with bad cohesion. The peripheral classes, on
the other hand, tends to become just data containers, doing nothing at all, which makes their
purpose unclear. A spider in the web design can normally be improved by moving some of
the peripheral classes further away from the spider class, as is done with Guest and Room in
figure 5.13b. This, improved, design does not have a spider class with references to all other
classes. Also, there is not a huge set of peripheral classes without references. Note that the
total number of references is the same in both designs in figure 5.13. Still, coupling is lowered
since the better design does not include a spider class with high coupling.

(a)

(b)

Figure 5.13 Two different designs of the same program:

(a) with high coupling (b) with low coupling

Just as is the case for high cohesion, low coupling is not something that can be achieved
once and for all. It is absolutely mandatory to always try to minimize the coupling. Also
parallel to high cohesion, low coupling does not apply only to classes, but to programming
constructs of all granularity, for example subsystems, packages and methods.

49
 Chapter 5 Design

5.3 Architecture
The architecture gives the big picture of the sys-
tem under development, it shows how the system
is divided into subsystems. It tells which prob-
lems the system can solve, and where in the sys-
tem each problem is solved. It does not, however,
tell exactly how the problem is solved, that be-
longs to design and coding. As an analogy, con-
sider the architectural plan of a building in figure
5.14. It ensures that the problem of moving be-
tween floors can be solved, since there is a stair.
It does not tell exactly how to construct the stair,
which materials to use, etc. And it most certainly
Figure 5.14 An architectural plan does not show is not a real, usable, stair. It is just a plan. Sim-
any details of the construction. Image by Gunnar ilarly, an architectural plan of a software system
Way-Matthiesen (stockholmskällan) [Public domain], via
Wikimedia Commons could ensure that for example data storage can be
handled, by including a database and a class or
package that calls the database. However, the architecture of the software system would not
be a detailed design of the database or the calling package, and it would definitely not be an
actual database or program, but instead a UML diagram or something similar.

Patterns
This section will cover architectural patterns, but first, let us make clear what a pattern is.
A pattern is a common and proven solution to a reoccurring problem. Typically, developers
realize that a particular problem in software development is solved many times, in different
programs, but the solution is always more or less the same. If this solution works well, it is
worth creating a formalized description covering the problem, variants of the solution, advan-
tages and disadvantages of the solution, etc. This formalized description is a pattern. If it
concerns architecture, it is an architectural pattern, if it concerns design it is a design pattern,
and so on. A collection of patterns is like a cookbook for software development. Knowledge
of patterns becomes a common vocabulary for software developers, that can be used to discuss
possible ways to solve a particular problem.

Packages and Package Private Visibility

Now that we are about to divide the system into
smaller subsystems, it is important to start using
packages and package private visibility. This is
how logical parts of the program are represented
in the Java language, without it, the division into
subsystems exists only in the minds of the pro-
Figure 5.15 Package private visibility.
grammers. Package private visibility means that

50
 Chapter 5 Design

a particular declaration (field, method, class, etc) is visible only to code in the same package
as that declaration. In UML, it is illustrated with the tilde character, see figure 5.15. In Java, it
is declared by omitting visibility modifier, do not write neither public, nor private (or any-
thing else). See appendix C.11, showing the implementation of figure 5.15. Note that package
private visibility is closely related to private visibility. Both are part of the implementation
and impose a strong limit on the visibility. Both make it possible to tell exactly which code
can see the declaration.

The MVC (Model-View-Controller) Architectural Pattern

The architectural pattern MVC (Model-View-Controller) tells that the system must be divided
into the subsystems Model, View and Controller, to avoid mixing code doing completely dif-
ferent things. Without such a division into subsystems, it would easily happen to mix user
interface code and business logic code in the same method. Say that we are coding, for ex-
ample, a bank account. A straightforward solution is to have a class Account that has a
field balance and methods deposit and withdraw. That is fine, such a class contains
only business logic (the actual functionality of the program) and its current state (the val-
ues of all variables). However, we also want to present the state of the account, for ex-
ample its balance, to the user. Therefore, it might seem adequate to add code handling for
example a HTML user interface to the Account class. This, however, would be a disas-
ter! HTML user interfaces and business rules for withdrawing money are two completely
different things. Mixing them just because they both use the same data, namely the ac-
count balance, would lead to extremely low cohesion and high coupling. Low cohesion
because the very same method would handle such
different things as UI and business logic, high
coupling because UI code and business logic code
would be inseparable, placed in the same method.
As a consequence, a HTML designer would have
to know Java and understand the business rules,
and a Java developer would have to understand
the web based user interface created with HTML
and CSS. Furthermore, it would be impossible to
reuse the HTML for other web pages, not to men-
tion the nightmare of changing to another user in-
Figure 5.16 The view and model packages. terface, or having multiple user interfaces to the
same program. Maybe a customer using the in-
ternet bank needs a web based UI and a bank clerk needs a UI of a Java program run locally.
To avoid such a disastrous mess, the MVC pattern tells us to create the subsystems view,
containing all code managing the user interface, and model, with the business logic code, see
figure 5.16.
Having separated the system into view and model, the two separate tasks user interface and
business logic are clearly separated into two subsystems, each with high cohesion. There is,
however, still a remaining problem. To understand it, let us first consider an analogy, namely
to build a new school. The school will be used by a large number of people in many different

51
 Chapter 5 Design

roles, for example students, teachers, headmasters, it

staff. All these may have ideas about the construc-
tion that they want to communicate to the construction
workers. There is also a large number of construction
workers in many roles, for example carpenters, elec-
tricians, plumbers. This is illustrated in figure 5.17,
which depicts two different organizations. The upper
organization, figure 5.17a, is quite chaotic since any-
one with an opinion about the construction is allowed
(a) to give input to any construction worker. It is easy
to understand that no usable school will ever be built
with such an organization. The lower image, figure
5.17b, on the other hand, illustrates a better organiza-
tion. Here, there is a steering committee (green per-
son) organizing the input. No-one talks directly with a
construction worker, instead all input goes to the steer-
ing committee person, who filters the input and de-
cides what to forward and to which worker.
The analogy to the MVC pattern is that the blue per-
sons represent classes in the view package, since they
(b) give input, and the orange persons represent classes
in the model, since they perform the desired work.
Figure 5.17 People involved in constructing The architecture depicted in figure 5.16 would lead
a new school. The orange persons sym-
bolize the construction workers (carpen-
to an organization of the software similar to figure
ters, electricians, plumbers, etc). The 5.17a. Any object in the view would call methods
blue persons symbolize people that will in any object in the model. Such a system would
use the school (teachers, students, admin- have very high coupling. A change to a class in the
istrative staff, etc), and therefore give di- model could affect any class in the view. Also, to
rectives to the construction workers.
(a) Chaotic organization without steering
change or update the view would be very difficult
committee. since it would not be clear how replacing or chang-
(b) Well-functioning organization with ing a class in the view would affect the work actually
steering committee (green person). performed in the model. The solution is to introduce
a class (or some classes) corresponding to the steer-
ing committee. Such a class is called a controller and is placed in the third layer of the
MVC pattern, the controller layer. Figure 5.18 shows all three MVC layers and the
Controller class. The controller shall contain the system operations, that is, the oper-
ations in the system sequence diagram made during analysis. A user action, for example
to click a button in the user interface, will result in one call from an object in the view
to a system operation in the controller. That operation in the controller shall call the cor-
rect methods in the correct objects in the model, in the correct order. This way, the work is
done in the model and it is the controller’s responsibility to know which object in the model
does what. The view will not have any knowledge about the model or dependence on it.

52
 Chapter 5 Design

To summarize, the MVC pattern tells us to di-

vide the system into three subsystems. The first
is view, which is responsible for presenting the
user interface and for interpreting the user’s in-
put. There must not be any code related to any
kind of user interface outside the view. The sec-
ond subsystem is controller. As stated above,
the controller’s responsibility is to call the correct
methods in the correct objects in the model, in
the correct order. The last subsystem is model,
which contains the program’s representation of
real world entities and is responsible for the actual
functionality of the system, the business logic.
Figure 5.19 is a sequence diagram showing the
flow from view, via controller, to model.
The advantages of the MVC pattern are that each
subsystem has high cohesion, and that there is low
coupling between user interface code and busi-
ness logic code since they are separated in dif-
Figure 5.18 The MVC layers and the
Controller class.
ferent subsystems. The view and the model can
now be developed separately, by different teams.
It is in fact possible to completely replace the view, or to have multiple views simultaneously,
without affecting the model in any way.

Figure 5.19 Method calls from view, via controller, to model.

Before leaving the MVC pattern, it is worth considering interaction between the three sub-
systems a bit more. Regarding view and controller, a remaining question is who handles flow
control between views. Suppose for example the the user interface shows a list with summary
information about different items. If the user clicks an item, a new view shall be displayed,
with detailed information about the clicked item. Which object knows that the list view shall
be replaced by the detailed view? Flow control is the responsibility of the controller, not
the view, but the controller does not know which view is currently displayed. The answer

53
 Chapter 5 Design

is that managing flow control between views is a task complex enough to give low cohesion
to whatever class it is placed in. It is best to introduce a new class, which has exactly this
responsibility. This object could be placed in the controller layer, but note that it is not a
Controller class.
Regarding communication between view and model, it is best that, as in figures 5.18 and
5.19, there is no such communication at all. That makes these two layers are completely
independent, reducing coupling as much as possible. If so, the only way to send data from
model to view, for showing to the user, is as return values to method calls from view, via
controller, to model. If that is possible, everything is fine. Unfortunately, it is often not
possible, since one method call might require many different return values. It might also be
that a view shall be updated when no call to the model has been made, for example as a result
of the model being updated by a call from another program, or because of a timer updating the
model regularly. An option could be to add lots of getter methods to the model, and let the view
use those to retrieve the required data. This solution has some big disadvantages, for example
that corresponding getter methods must be added
to the controller, which will make it terribly
bloated and messy. Also, the view can not know
exactly when to call those getters, since it can not
know when the model changes state, if the state
change is not initiated by the view itself. There
is an elegant solution to this problem, that will be
covered later. For now, all considered scenarios
will allow data to be passed from model to view
as return values to method calls via controller.

The Layer Architectural Pattern

The Layer architectural pattern is more general
than the MVC pattern. While MVC concerns the
model, view and controller layers in particular,
the layer pattern just says that the system shall be
divided in layers. MVC solves the problem that
user interface and business logic risk to be mixed.
Layer applies the same reasoning, but to any two
different kinds of code. Just as mixing user inter-
face and business logic brings low cohesion and
high coupling, so does mixing for example busi-
ness logic and database calls. Calling a database
is a separate task, in no way related to perform- Figure 5.20 Often used layers.
ing business logic on objects in the model. This
means there shall be a separate layer dedicated to database calls. Continuing this reasoning, it
is important to always be prepared to add a new layer. That must be done whenever writing
code that will give low cohesion to whatever existing layer it is placed in. As an example,
consider the main method. Its task is to start the program, which is not related to any of the

54
 Chapter 5 Design

layers that have been mentioned so far. Therefore, yet a new layer must be introduced, whose
responsibility is to start the application.
Exactly which layers there shall be in a system is a matter of discussion, and it also differs
from system to system. However, all layers that have been mentioned here are often present.
Those layers are depicted in figure 5.20, they are view, controller, model, dbhandler
(sometimes called integration, responsible for calling the database), data (the actual
database) and startup, which includes the main method and all other code required to start
the application.
Always strive to keep dependencies in the direction illustrated in figure 5.20, that is from
higher (closer to the user) layers to lower (further from the user) layers. Those dependencies
are unavoidable, since execution is initiated by the user, and to call lower layers there must
be dependencies. Dependencies in the opposite direction are, however, not needed. There is
nothing forcing lower layers to call higher layers. Since such dependencies are unnecessary,
introducing them means unnecessarily high coupling. Also, higher layers tend to be less stable
than lower layers. For example, it is more common to change user interface layout than it is
to change business rules, and yet less common than to change those rules is to remove entities
represented in the database.
To conclude, the layer pattern has important advantages. If layers are correctly designed,
they form subsystems with high cohesion and low coupling. Also encapsulation applies to
layers, the public interface of a layer shall be as small as possible, not revealing more than
required of the layers internal workings. When encapsulation, cohesion and coupling are used
to make layers independent, it becomes easy to maintain the layers and to divide development
of different layers between different developers or teams of developers. It is also easy to reuse
code, since a layer can provide a well-designed public interface, callable from any code in a
higher layer.

The DTO (Data Transfer Object) Design Pattern

As the number of layers increase, so does the need to pass data between layers. This often
leads to long parameter lists in many methods as data is passed through the layers. Consider
for example registering a new user in some community. Say that registration means to enter
name, street address, zip code, city, country, phone number and email address. These are
seven string parameters that shall be passed though all layers from user interface to database,
which means there will be (at least) method declarations similar to those in listing 5.4.

1 //In the controller layer

2 public void registerUser(String name, String streetAddress,
3 String zipCode, String city, String country,
4 String phone, String email) {
5 //Call to model
6 }
7
8 //In the model layer
9 public void registerUser(String name, String streetAddress,

55
 Chapter 5 Design

10 String zipCode, String city, String country,

11 String phone, String email) {
12 //Call to dbhandler
13 }
14
15 //In the dbhandler layer
16 public void registerUser(String name, String streetAddress,
17 String zipCode, String city, String country,
18 String phone, String email) {
19 //Call to database
20 }

Listing 5.4 The same method signature often appears in many different layers. This is problematic if the method
has a long parameter list.

Just to make many method calls is not a problem, but the long parameter list is. First, it is
difficult to remember the meaning of each parameter, especially when they all have the same
type, as is the case here. Second, a long parameter list means a large public interface, and
thereby a bigger risk that it is changed. An often used method to get rid of the parameter list
is to use a data transfer object, DTO. Such an object is a just data container, without any logic.
Its only purpose is to group data in the same class, see listing 5.5.

1
2 //The DTO
3 public class UserDTO {
4 private String name;
5 private String streetAddress;
6 private String zipCode;
7 private String city;
8 private String country;
9 private String phone;
10 private String email;
11
12 public UserDTO(String name, String streetAddress, String zipCode,
13 String city, String country, String phone,
14 String email) {
15 this.name = name;
16 this.streetAddress = streetAddress;
17 ...
18 }
19
20 public String getName() {
21 return name;
22 }
23
24 public String getStreetAddress() {
25 return streetAddress;

56
 Chapter 5 Design

26 }
27
28 ...
29 }
30
31 //In the controller layer
32 public void registerUser(UserDTO user) {
33 //Call to model
34 }
35
36 //In the model layer
37 public void registerUser(UserDTO user) {
38 //Call to dbhandler
39 }
40
41 //In the dbhandler layer
42 public void registerUser(UserDTO user) {
43 //Call to database
44 }

Listing 5.5 Here, the problematic parameter list of listing 5.4 has been removed by introducing a DTO

An obvious objection is that the long parameter list is not gone, it is just moved to the
constructor of the DTO, UserDTO, on lines 12-14 in listing 5.5. However, it now appears only
in one place. If it is changed, only one public interface is changed, not one in each layer. Also,
it is now obvious that all user related data belongs together.

5.4 A Design Method

Finally, all the theoretical background is covered. Having sufficient knowledge about UML
class, sequence, and communication diagrams; the design concepts encapsulation, cohesion
and coupling; and the architectural patterns MVC and layer, it is time to look at how to actually
design a program. This section describes a step-by-step method for design, which will be used
to design the RentCar case study in the next section.

1. Use the patterns MVC and layer. This means to create one package for each layer
that is supposed to be needed. Exactly which layers that is, is a matter of discussion.
An educated guess that is valid for many programs is to use the layers depicted in figure
5.20. The class Controller can also be created already now. Illustrate the packages
and Controller in a class diagram.

2. Design one system operation at a time. The system sequence diagram shall guide our
design, it shows exactly which input and output the program shall have. Also design
enough of the system’s start sequence (initiated by the main method) to be able to test
run the newly designed system operation. When deciding which classes and methods

57
 Chapter 5 Design

there shall be, use interaction diagrams. An interaction diagram shows the flow through
the program, how methods call each other. Whether to use sequence or communication
diagrams is a matter of taste.

3. Strive for encapsulation with a small, well-defined public interface, high cohesion
and low coupling. When adding new functionality, create the required methods in a
way that these goals are met to the highest reasonable degree. This is much easier said
than done, and often requires much thought and discussion. It helps to remember that an
operation shall be placed in a class representing the abstraction to which the operation is
associated, and that has the data required for the operation. The domain model helps to
find new classes that can be introduced. Also, always be prepared to change previously
designed system operations if that improves the overall design.
When designing, favor objects over primitive data and avoid static members, since nei-
ther primitive data nor static members are object oriented. When using these, the entire
object concept is completely ignored, and the prime tool (objects) to handle encapsula-
tion, cohesion and coupling is thrown away.

4. Maintain a class diagram with all classes. When done designing a system operation,
summarize the design in the class diagram created in bullet 1, which will contain all
classes in the program. Such a diagram gives a good overview.

5. Implement the new design in code. Design is not an up front activity that can be done
once and for all for the entire program. Instead, it shall be done in iterations, as soon as
a design is ready it shall be implemented in code, and thus evaluated. Here, this step is
postponed until the next chapter. The reason is that seminar two would otherwise be
far too big. Also, postponing programming allows focusing on design without getting
lost in code. However, when designing, it is important to have an understanding of how
the design can be implemented in code.

6. Start over from bullet 2 and design the next system operation.

5.5 Designing the RentCar Case Study

As an example, the RentCar case study is designed in this section. For convenience, the
specification, SSD and domain model are repeated below, in figures 5.21, 5.22 and 5.23. When
creating the design, be sure to have an understanding of how it may be implemented in code. If
that is not clear, now is the time to repeat Java programming, for example by reading chapter 1,
especially sections 1.2 and 1.3.

Step 1, Use the Patterns MVC and Layer

Following the method described in section 5.4, the first thing to do is to create a class diagram
with the anticipated layers. There is no reason to deviate from the typical architecture of figure

58
 Chapter 5 Design

1. The customer arrives and asks to rent a car.

2. The customer describes the desired car.
3. The cashier registers the customer’s wishes.
4. The program tells that such a car is available.
5. The cashier describes the car to the customer.
6. The customer agrees to rent the described car.
7. The cashier asks the customer for name and address, and also for the driving license.
8. The cashier registers the customer’s name, address and driving license number.
9. The cashier books the car.
10. The program registers that the car is rented by the customer.
11. The customer pays, using cash.
12. The cashier registers the amount payed by the customer.
13. The program prints a receipt and tells how much change the customer shall have.
14. The program updates the balance.
15. The customer receives receipt, change and car keys.
16. The customer leaves.
4a. The program tells that there is no such car available.
1. The cashier tells the customer that there is no matching car.
2. The customer specifies new wishes.
3. Execution continues from bullet three in basic flow.

Figure 5.21 The RentCar scenario.

Figure 5.22 The RentCar system sequence diagram.

59
 Chapter 5 Design

Figure 5.23 The RentCar domain model.

5.20. Therefore, after having introduced the Controller class, the design looks as in figure
5.24.

Figure 5.24 The first version of the RentCar design class diagram.

60
 Chapter 5 Design

Step 2, Design One System Operation at a Time

The view will not be designed here, instead there will be a single class, View, in the view
package. This class is a placeholder for the real view, which certainly would consist of more
than one class. This way, there is no need to bother about view technologies like window
management, console IO or HTTP. Also, as far as design is concerned, there is nothing con-
ceptually different in designing the view than there is in designing any other layer; exactly the
same reasoning as here is followed when the view is designed.
The system operations are designed in the order they are executed according to the SSD,
figure 5.22. The first operation in the SSD is searchMatchingCar, which takes the param-
eter searchedCar and returns the value foundCar. The first step is to create an interaction
diagram. Since the MVC pattern says the controller shall contain the system operations, the
method searchMatchingCar can be added to the controller right away. Here, however,
comes the first design decision. Which is the type of the parameter searchedCar and the
return value foundCar?

Step 3, Strive for encapsulation with a small, well-defined public

interface, high cohesion and low coupling
The question, which is the type of the parameter searchedCar and the return value foundCar,
is the first of a large number of design decisions, let’s consider it carefully. The answer shall be
guided by the concepts encapsulation, cohesion and coupling. The purpose of searchedCar
is to represent the customer’s requirements on the car to rent, and the purpose of foundCar
is to describe the available car that best matches those requirements. The design currently
contains only two classes, View and Controller. Obviously, it would be lousy cohesion to
let any of those represent the customer’s requirements or the matching car. The features of
a car is not just one value, but a quite large set. The requirements specification, figure 5.21,
does not tell exactly which features the customer can specify. This is definitely something
we would have to ask about if the program should be used for real. In this exercise, we have
to decide on our own, let’s say that the customer can wish for price, size, air condition, four
wheel drive and/or color. These values clearly belong together, as they describe the same ab-
straction, a car. Therefore, they should be fields in the same class, which must be introduced
to the design. Can the domain model, figure 5.23, give any inspiration about this new class?
Yes, it shows the class Car, which seems to be an important abstraction since it has many
associations. The Car class can also be used for the return value, foundCar. This object,
however, represents a specific car, not just any car matching the specified criteria. Therefore,
the registration number must also be a field in this class. Now, having decided on the rep-
resentation of searchedCar and foundCar, it is possible to create the first version of the
searchMatchingCar interaction diagram, figure 5.25.
There are a few things worth noting in this diagram. First, the name searchedCar appears
in three different places, the object name, method call one and method call two. The fact
that the same name is used implies that it is in fact the very same object in all three places,
which is important information for the reader. Second, data can not appear out of nothing,
since searchedCar is a parameter in method call two, it must be clear from where this object

61
 Chapter 5 Design

Figure 5.25 The first version of the searchMatchingCar design interaction diagram.

comes. This is illustrated in method call one, where it is created. But what is the origin of the
parameters in method call one? That is explained in the note, they are entered by the user in
the user interface. Last, the diagram does not tell in which layer the Car class is located. It is
quite OK not to show layers in the interaction diagram, but we still must decide the location
of Car. In fact, all layers are candidates, since it already appears in view and controller,
and we can guess that it will be passed through model to dbhandler, since a search in the
database for a matching car is probably required. A rule of thumb is to place a class in the
lowest layer where it is used, in order to avoid dependencies from lower to higher layers. This
would indicate that Car should be placed in dbhandler. However, there are other questions
as well, is Car a DTO or is it the actual model object, the entity, with the business logic?
Also, does the entity (in the model) contain any business logic at all, or has it only got getter
methods? In the latter case, there is no real need to introduce both a DTO and an entity, since
they would be identical. Instead, the Car class in the model could be considered to be a DTO.
These questions can not be answered until more of the system is designed. For now, we just
choose the simplest solution, namely to let Car be a DTO, place it in dbhandler, and not
add an entity. This decision might have to be changed later. Note that if we had decided to
turn Car into an entity object, it could no longer have been created by the view, since model
objects shall only be called by the controller. Let’s change the name to CarDTO to make clear
it is a DTO, this makes the communication diagram look as in figure 5.26.

Figure 5.26 Car is renamed to CarDTO.

62
 Chapter 5 Design

Next, it is time to decide which object is called by Controller. Shall some object in the
model be created or shall the controller just make a search in the database? The answer is that
a model object shall be created if it is of any use after this call. For example if it later shall be
stored in the database or if it will be used in a future system operation. As far as we know now,
none of these cases apply, there is no future use for a model object representing the search.
Therefore, Controller will just call an object responsible for searching in the database. This
object, let’s call it CarRegistry, will reside in the dbhandler layer, since the purpose of
that layer is exactly this, to call the database. The database itself, represented by the data
layer, would normally be another system, called by the CarRegistry class. Here, however,
there is no database. Thus, instead of calling the database, CarRegistry will just look in an
array of available cars. The final design of searchMatchingCar is in figure 5.27. Note that
this diagram has no notion at all of the loop around the system operation, that is drawn in the
system sequence diagram. This is because the loop condition, noMatchingCar is not visible
in the program, but is completely decided by the cashier (after talking with the customer).
Exactly the same flow, depicted in figure 5.27, is executed again and again, until the customer
is satisfied.

Figure 5.27 Call to dbhandler layer is added.

The next task is to design as much of

the start sequence, in the main method, as
is needed to run the newly designed sys-
tem operation. That start sequence must
create all objects that are not created dur-
ing execution of the system operation it-
self. There exists currently four objects
in total, searchedCar and nameless ob-
jects of the classes View, Controller
and CarRegistry. Of these, only
searchedCar is created in the design of
Figure 5.28 The start sequence in the main method.
the system operation, the other three must be

63
 Chapter 5 Design

created by main. But not only must they be created, they must also be given references to
each other to be able to call each other. In particular, View calls Controller and must there-
fore have a reference to Controller. Also, Controller calls CarRegistry and must
thus have a reference to CarRegistry. One option is that main creates all three objects and
passes references as needed. Another option is that main creates fewer objects, for example
only View, which in turn creates Controller, which finally creates CarRegistry. Let’s
chose the former alternative, main creates all three objects, see figure 5.28. The other solution
could be problematic if in the future there is the need to create for example a Controller
without a CarRegistry. This might also indicate that the controller gets low cohesion if it
creates CarRegistry.

Step 4, Maintain a Class Diagram With All Classes

Figure 5.29 RentCar design class diagram after designing searchMatchingCar.

The last task in the design of searchMatchingCar is to summarize what has been done in

64
 Chapter 5 Design

a class diagram, see figure 5.29. Note that it is not mandatory to include all attributes, methods
and references if they do not add any important information, but only obscure the diagram.
For example, it is common not to include references to DTOs, since they are used in many
different layers and are considered as data types. Not including references to DTOs is similar
to not including references to java.lang.String, which can also be considered a data type.
Before leaving searchMatchingCar, we evaluate it according to the criteria encapsula-
tion, cohesion and coupling. To start with encapsulation, all methods are public. This is not
exactly an ideal situation, but often quite unavoidable early in the design. We are creating the
different layers and tying them together. In fact, all methods are called across layer borders,
and it has to be that way. Otherwise, it would not be possible to communicate between lay-
ers, since there are still very few methods. All fields, on the other hand, are private, which
is very good. Looking at cohesion, we can safely say that all classes do what they are meant
for, nothing more. Main starts the program, View just calls the controller, Controller has
a system operation that calls a search method in the CarRegistry. CarRegistry has only
this search method and CarDTO, finally, has no methods at all. Regarding coupling, there is
a chain of dependencies from higher to lower layers, that is from view to controller to
dbhandler, which is exactly the purpose of the layer pattern. Also, it is perfectly in order
to have dependencies from Main to the other layers, since the task of Main is to start the
other layers. It would most likely not be appropriate if Main had references to many classes
in the same layer, that would probably be too high coupling. For the moment, however, Main
references only one class in each layer.
That concludes the design of the searchMatchingCar system operation. The question
naturally arises, whether all this designing is really necessary just to fetch an element from
an array? The answer is yes, most definitely yes. A professional programmer should, and
normally does, make this kind of considerations all the time. However, having gained more
experience by designing more programs, the reasoning made in this section can often be done
quite quickly.

The registerCustomer System Operation

The next system operation is registerCustomer. The first thing to do is to introduce the
system operation as a method in the controller, since all system operations shall appear as
controller methods. What objects shall the registerCustomer method call? This is the
time to introduce model objects, since the result of customer registration is needed in future
system calls, for example when a rented car is associated with the renting customer. It seems
quite natural to add a Customer class, there is also such a class in the domain model. Again
comes the same consideration as for the Car class, is this a DTO or an entity or is there no
difference between the two? That question can not be properly answered until more is known
about the program. For the moment, we choose an easy solution and treat it as we treated
the car object. Make the class a DTO, place it in the lowest layer where it is used, namely
model, and create the object in view. Treating car and customer the same way should
also make it easier to understand the program. We must, however, be aware that these choices
might have to be changed later on, as designing proceeds. Having solved this problem, at
least temporarily, another question immediately appears. According to both domain model

65
 Chapter 5 Design

and SSD, CustomerDTO has the attributes name, address and drivingLicense. Shall
these be strings or new classes? In order to shorten the discussion, the domain model is
followed without further consideration. This means name becomes a string, while the other
two becomes new classes, AddressDTO and DrivingLicenseDTO. Now it is possible to
draw a UML diagram illustrating the call of the registerCustomer method, figure 5.30.

Figure 5.30 The customer object and the registerCustomer method.

We are not done yet, sending a DTO to the controller does not serve any purpose on its own.
Remember that a DTO shall be considered a data type, like int or String. To complete
customer registration, customer data should reasonably be stored somewhere in the model
(or database). This is a good time to add the Rental class, which is a central class in the
domain model. High cohesion is achieved by letting a rental object represent the entire
rental transaction. This object will know which customer performed the rental, which car was
rented, and other facts related to the rental, by having references to appropriate other objects.
The final design of the registerCustomer system operation look as in figure 5.31.
There is no need to add anything to the main method, since all new objects that were
introduced in this system operation are created in the interaction diagram in figure 5.31. These
objects are address, drivingLicense, customer and the unnamed Rental object.
A word of caution before proceeding to the next system operation. There are now three
different DTOs stored in the model, in the Rental object, and there will likely be even more
as we proceed. We have not considered how these are handled in the model. Are they simply
kept or is all data copied to some other object? This question will be left unanswered until the
design is implemented in code. However, it is a problem that the DTO objects are referenced,
and therefore potentially updated, by both view and model. Once view has passed a DTO as
parameter to controller, it should never be updated again by view. To make sure this does
not happen, all fields, and also the classes themselves, shall be final. This makes the DTO

66
 Chapter 5 Design

immutable, which means none of its fields can ever change value. There is no UML symbol
for this, it is illustrated with a note in the class diagram, figure 5.32.

Figure 5.31 The Rental class symbolizes the entire rental transaction.

The bookCar System Operation

The purpose of the bookCar system operation is to register which car will be rented. Note
bullet ten in the specification, The program registers that the car is rented by the customer.
This is not illustrated in the system sequence diagram. Correctly, since it is an operation
internal in the system, but still it must appear in the design.
The name of the parameter that specifies the car in the SSD is foundCar, which is the same
name as the variable that was returned from the searchMatchingCar system operation. This
indicates that these two are the same object. This information will be kept in the design, by
using the same name, foundCar, for both objects also here.
The foundCar object must be passed to the rental object in the model, to be associated
with the current rental and thereby the current customer. That is not enough however, the car
database must also be updated to show that the car is now booked, to prevent other customers
from renting it. This raises the question, what to store in the database? Just the fact that
the car is now booked, or all information in the entire rental object? The latter must be the
correct choice, otherwise the information about this rental will be lost when the next rental is
initiated. Of course, in a real project, this would be discussed with a domain expert (someone
working at the car rental company), but here we have to decide on our own. The decision to
store all rental information creates a new problem, shall a rental be stored in a database named
carRegistry? Isn’t that name a bit misleading? Either the name carRegistry must be

67
 Chapter 5 Design

Figure 5.32 RentCar design class diagram after designing registerCustomer.

68
 Chapter 5 Design

changed, or a new data store must be created. Let’s try to get inspiration from the domain
model, it shows a RentalCatalog and a CarCatalog. This indicates that there should be
different data stores for cars and rentals. It can also be argued that separating these two classes
creates higher cohesion. However, these two are not exactly what we are looking for, in the
DM they contain specifications of rentals and cars, but in the design we are handling particular
instances of rentals and cars. Also, comparing the DM and the design diagrams, it becomes
clear that the design so far contains no rental or car specification stores, is that a problem? This
is again something that should be discussed with the domain expert. But let’s not deviate to
much from the SSD we are implementing now, we will not consider car or rental specifications
here, since they are not mentioned in the specification.

Figure 5.33 The bookCar design interaction diagram.

With the cohesion argument, a RentalRegistry is added, this gives higher cohesion than
storing rentals and cars in the same data store. This results in the design in figure 5.33. Re-
member that CarRegistry and RentalRegistry are not the actual databases, but classes
calling the database. There is nothing stopping us from letting both those call the same under-
lying database if that would be appropriate. In this course, we do not implement any database,
so we do not have to consider that problem.
Note that the CarRegistry method that marks the car as beeing booked is called setState
OfBookedCar, and not bookCar. The reason is that booking the car is business logic, which
belongs in the model. The purpose of the integration layer is to access the data store, not to
perform business logic, like booking a car. Therefore, this method just saves a booked state,
without caring about eventual business rules related to performing a booking.
It is a quite interesting decision to let Rental, instead of Controller call setStateOf
BookedCar. The motive is that the controller should not have detailed knowledge about all
details of all system operations. That would lead towards spider-in-the-web design, with the
controller as spider. Now that calls to registers are being made from Rental, it might be
adequate to move more register calls from Controller to Rental. Then there would of
course be the risk that, in the end, Controller does nothing but forward calls to Rental,
which then becomes the spider class. This reasoning is not an unimportant academic exercise,
it is quite common design problems both to have a controller doing all work itself, and to
have a controller doing nothing but forwarding method calls to another class. To shorten this
discussion a bit, the design is kept as in figure 5.33.

69
 Chapter 5 Design

The RentalRegistry object is not created in any design diagram, it must therefore be
created when the system is started. Figure 5.34 shows program startup with instantiation of
RentalRegistry added. With this modification, main creates two objects in the dbhandler
layer. This is a warning sign that it might be getting unnecessarily high coupling to that layer.
Also, the dbhandler layer might have a bit bad encapsulation, since it has to reveal the ex-
istence of CarRegistry and RentalRegistry to main. These problems can be solved by
changing that startup design to the one in figure 5.35, where the class RegistryCreator is
responsible for creating the registries, and thereby hides their existence to main. This solution
will be discussed and improved further in chapter 9. For now, we conclude that the design
in any of figures 5.34 or 5.35 can be used, since the problem regarding encapsulation in the
dbhandler layer is not yet very big. But it might grow in the future, if more registries are
added.

Figure 5.34 The start sequence in the main method.

The design class diagram, figure 5.36, is now becoming quite big. In order to reduce it, the
DTOs are omitted. Another option would have been to split it into more, smaller diagrams.
This class diagram illustrates the start sequence in figure 5.35, not 5.34. Note that the con-
structors of CarRegistry and RentalRegistry are package private, since they are called
only by RegistryCreator, which is located in the same package as those registries.

70
 Chapter 5 Design

Figure 5.35 The start sequence when RegistryCreator is added.

Figure 5.36 RentCar design class diagram after designing bookCar.

71
 Chapter 5 Design

The pay System Operation

Only cash payment is implemented in the current
iteration. Just as is the case for any system opera-
tion, there will be a method in the controller with
the same signature as the operation in the SSD,
that is void pay(amount). What is the type of
the parameter amount? So far, int has been used Figure 5.37 The pay system operation.
to represent amounts, for example the price of a
rental. Looking in the domain model, however, there is a class called Amount that represents
an amount of money. It is most likely a good idea to change the design to use that type instead.
Generally, it is a bit dangerous to force an amount to have a specific primitive type. For ex-
ample it is not clear whether an amount can have decimals or not. By introducing the Amount
class, the primitive type of the amount is encapsulated inside that class, and can thus easily be
changed. It is a great joy to see that the introduction of Amount only requires changes from
int to Amount in one single class, CarDTO. This is due to the encapsulation of car properties
in CarDTO. The pay interaction diagram now looks as in figure 5.37.

Figure 5.38 Payment and CashRegister handling the pay system operation.

Which object in the model shall handle a payment? Rental is the only class that can come
under consideration without completely ruining cohesion. Is it reasonable that Rental shall
prepare the receipt and perform the task listed in bullet 14 in the scenario, namely to update
the balance? The answer must be no, Rental represents a particular rent transaction, it is not
responsible for receipt creation or for maintaining the balance of the cashier’s cash register.
This means a new class must be created. The DM has no really good candidate for this class,
which probably means something was missed when it was created. Possible classes in the DM
are Payment and Cashier. The former is associated to Receipt, which in turn is associated
to Change, and seems to be a good candidate for handling one specific payment. One specific
payment is, however, not related to the balance in a cash register. Therefore, Payment shall
not handle the balance. Looking in the DM, Cashier is the only possible candidate for

72
 Chapter 5 Design

handling the balance, but is a balance really an attribute of the cashier that worked at the cash
register where the balance was generated? The answer must be no, which means none of
the classes in the DM can be used. The most reasonable solution seems to be to introduce a
new class CashRegister, representing the cash register that has the particular balance. The
pay design, after introducing the Payment and CashRegister, is depicted in figure 5.38.
The payment class is called CashPayment instead of Payment, because we are anticipating
that future handling of credit card payments will be quite different and therefore be placed
in a different class. The rental that is being paid is passed to calculateCost, call 1.2.1,
since CashPayment will have to ask the payment about information when calculating the
total rental cost.

Figure 5.39 Receipt printout is added.

As a result of pay, a receipt shall be printed on the printer, which, according to the SSD,
is an external system. Calling an external system is normally handled by a specific class,
which represents the external system and handles all communication with it. This class can
be named after the external system, here, it will be called Printer. This class is not the
actual printer, but a representation of the printer in the program being developed. In which
layer shall this class be placed? Actually, it does not fit in any of the current layers without
giving bad cohesion to the layer where it is placed. There are two main options, either to
create a new layer or to extend (and rename) dbhandler to handle interaction with any other
system, so far databases and printers. The former seems like a road to fragmentation of the
system into many small layers, to high coupling with many references, and to less possibility
for encapsulation, given the many small units. The latter option seems to be a road to low
cohesion in dbhandler. With the current knowledge about the system, it is quite impossible
to tell which option is the best. More or less by chance, the latter option is chosen and the
dbhandler layer is renamed to integration, which is a relatively commonly used name

73
 Chapter 5 Design

for a layer responsible for interaction with external systems. The resulting pay design can
be seen in figure 5.39. This design is a bit underspecified, for example it is not clear exactly
how Receipt will gather the receipt information. Actually, it is not even clear exactly which
information the receipt shall contain. However, it is clear that what is designed is sufficient to
allow Receipt to gather the information from Rental. The remaining details will be decided
when the design is implemented in code.
Two objects were introduced without being created, namely the Printer and CashRegister
objects, which must therefore be created during startup. Shall Printer be created by Registry
Creator (which must then be renamed), or shall it be created directly by main? Let’s not
include it in RegistryCreator, since, after all, a printer connection is completely different
from a database connection. The RegsitryCreator will be responsible only for connecting
to the database. Perhaps the same connection can be used for both the car and rental registries,
but most certainly not for the printer. This decision gives the final startup design of figure
5.40.

Figure 5.40 The complete startup design.

Why is the CashRegister object created by Controller, when all other objects are
created by main and sent to Controller? This is a trade-off between two contradicting
arguments. On one hand, coupling is lowered if main is not associated to all other objects cre-
ated during startup. On the other hand, cohesion of the Controller constructor is increased
if it does not create loads of other objects, besides controller. The design of figure 5.40
balances these arguments. Since Controller is the entry point to model, it makes sense that

74
 Chapter 5 Design

it creates the model objects, like CashRegister. The main method creates central objects
in the integration, controller and view layers, but nothing more.
The last thing to do is to draw the design class diagram, which can be seen in figure 5.41.
Note that there is no data layer, it is not needed since there is no database.

Figure 5.41 The class diagram after pay is designed.

75
 Chapter 5 Design

Evaluate the Completed RentCar Design

Before leaving the design, it should be evaluated according to the concepts encapsulation,
cohesion, and coupling. Starting with encapsulation, is there any method, class, package or
layer that has bad encapsulation? Bad encapsulation means that there is a member with too
high visibility, public instead of package private or private. With the current design, there is
no visibility that can be lowered. Still, there are very few methods that are not public, which
is not the desired result. Also, there is quite high coupling. For example Controller is
associated with all classes in model and most classes in integration. This situation would
be improved if model classes called each other instead of passing through controller. One
thing that can be done is to change the Rental method Receipt getReceipt() to void
printReceipt(Printer printer). This moves the printer call from Controller to
Rental, removes the association from Controller to Receipt and makes the Printer
constructor package private. This gives the pay design in figure 5.42 class diagram in figure
5.43. This is at least a bit better. Further improvements could be considered, for example
to somehow let CashPayment call CashRegister (or vice versa). This would remove one
more association from controller. However, the current design is quite acceptable, let’s
leave it like that. It is normal that early in development, a large part of the created members
belong to the public interface. The last thing to do is to consider cohesion. There seems to
be no issues related to cohesion, all layers, classes and methods do what they shall, nothing
more.

Figure 5.42 Improved pay design, with less coupling from Controller.

76
 Chapter 5 Design

Figure 5.43 The final design class diagram.

77
 Chapter 5 Design

5.6 Common Mistakes

Below follows a list of common design mistakes.

• The design has a spider-in-the-web class, which is often the controller. The solution is
to remove associations between the spider and some peripheral classes, and instead add
associations between peripheral classes. This has been covered in detail previously.

• Objects are not used sufficiently, instead primitive data is passed in method calls. It is
not forbidden to use primitive data, but always consider introducing objects, especially
if there are long lists of parameters in a method or attributes in a class. Not using objects
means that the prime tool (objects) to handle encapsulation, cohesion and coupling is
thrown away.

• There are unwarranted static methods or fields. It is not forbidden to use static mem-
bers, but there must be a very god reason why there are such. Static members do not
belong to any object, and are therefore, just as is the case for primitive members, us-
ing them means that the prime tool (objects) to handle encapsulation, cohesion and
coupling is thrown away.

• There a too few classes. It is of course very difficult to tell how many classes there
should be in a certain design, but in some cases there are clearly too few. An example
is if the model consists of only one class, which performs all business logic. Cohesion
is the prime criteria used to decide if there is a sufficient number of classes, too few NO!
classes normally means that some existing class(es) has low cohesion.

• Too few layers is perhaps a less disastrous mistake than too few classes, especially
early in the development, when the program is relatively small. Still, if one of the
layers view, controller, model, startup or integration is missing, there
must be a reason why that is the case. If one or more of those layers has another name
is probably no problem, what matters is that they exist.

• There can also be too few methods. This can be discovered by evaluating if existing
methods have high cohesion. A method should have one specific task, explained by
its name. However, there can be too few methods even if all existing methods do have
high cohesion. This is the case if some of the program’s tasks is simply not performed
in any method. If so, the design is not complete and a new method must be introduced,
performing the missing task.

• The MVC pattern might be used the wrong way. Under no circumstance must there be
any form of input or output outside the view.

• Also the layer pattern might be used the wrong way. All layers must have high cohesion
and there should be calls only from higher (closer to the user) layers to lower layers.

78
 Chapter 5 Design

• Data appears out of nothing. Always consider if it is possible to implement the design
in code. That is not possible if a certain variable is passed in a method call, but the
variable does not exist in the calling method.

• The class diagram is too big and is therefore unreadable. The diagram might be un-
readable because it is messy, showing many details, or because it has been shrunk to fit
on a printed page, making the text too small. The solution is to either split it in more, NO!
smaller, diagrams, or to remove details that are not needed to understand the design.
Examples of things to remove are DTOs, private members and/or attributes. Remem-
ber that the goal of removing details is to make it easier to understand the diagram, do
not remove things that are required for understanding. After having removed details, it
might become difficult to use the class diagram as a template when coding. However,
there should still be a complete design of each class in the design tool, this can be used
when programming.

79
Chapter 6

Programming
This chapter describes how to implement the design in code, and how to test the code. Coding
is never a straightforward translation of the design diagrams. Most likely there are coding
details that were not considered during design. It is also quite likely to discover actual design
mistakes. This means there is no sharp line between the design and implementation activities,
there will be need for decisions regarding program structure, encapsulation, cohesion, cou-
pling and so on also when coding. In addition to this, there will also be questions regarding
code quality, that are not really design issues, for example how to name variables and how to
write comments in the code.
As soon as a certain functionality is implemented in code, it should be tested. This chapter
covers fundamentals of unit testing. This includes guidelines about what, when and how to
test, and an introduction to technologies that facilitate testing.

6.1 Dividing the Code in Packages

A package name consists of components, separated by dots. The first components shall always
be the reversed internet domain of the organization, e.g., se.kth. This is to avoid name con-
flicts with packages created by other organizations. Following that, there are normally com-
ponents that uniquely identifies the product within the organization, e.g., department and/or
project names, iv1201.carRental. Finally, there are the components that identify a par-
ticular package within the product. This part could start with layer name, e.g., model. If the
layer is large, a single package containing everything in that layer might get low cohesion.
If so, the package can be divided according to functionality, e.g., payment. When following
all these rules and guidelines, a package in the model of the car rental application, handling
payment, shall be named se.kth.iv1201.carRental.model.payment
Sometimes, a class does not clearly belong to a specific package, but is needed in many
different packages. This behavior of a class gives important information, often that the class
does not really fit in the layer structure of the program. This might be because of a design
mistake, but another common reason is that the class is a utility class. These are normally
not application specific, but provide some kind of general service, for example string parsing
or file handling. Such utility classes are often placed in a package that does not belong to a
specific layer, but is instead called for example util. The full name of that package in the car
rental application would be se.kth.iv1201.carRental.util.

80
 Chapter 6 Programming

6.2 Code Conventions

It is essential that code is easy to understand, since it will, most likely, be read and changed by
many more developers than the original creator. To make the code easy to understand, every-
one must agree on a set of rules for formatting, naming, commenting, etc. These rules form
a code convention. Originally, there was a Java code convention published by Sun Microsys-
tems. It is no longer maintained by Oracle, but is still available at [JCC]. A good summary
of Java coding standards, which is close to the original code convention, is available at [JCS].
In addition to these documents, organizations that produce code often have their own code
convention. It is essential to agree on which code convention to follow in a particular project.
Below, in table 6.1, follows a brief summary on very frequently used naming conventions
for Java. Note, however, that a full code convention is much more extensive than these short
rules. It is a good idea to read through one of the documents mentioned above.

Name Description Example

Type
package First letter of each part lowercase. Start with se.kth.iv1201.
reversed internet domain, continue with product rentCar.model or
name and end with unique package name se.kth.iv1201.
rentcar.model
class and Full description, first letter of each word CashRegister
interface uppercase
method Full description, first letter lowercase, first letter calculateTotalCost
of non-initial words uppercase.
variable, Full description, first letter lowercase, first letter paidRental
field, of non-initial words uppercase.
parameter
constant This applies to final static fields. Only MILES_PER_KM
uppercase letters, words separated by underscores.

Table 6.1 A few, very commonly used, Java naming conventions.

6.3 Comments
There should be one javadoc comment for each declaration that belongs to a public interface.
Javadoc comments start with /** and end with */. These are used to generate html files with
api documentation using the javadoc jdk command. Most IDEs (for example NetBeans and
Eclipse) provide a graphical user interface to the javadoc command.

81
 Chapter 6 Programming

The javadoc comment shall describe what the commented unit does, but not how it is done.
How belongs to the implementation, not to the public interface. Method comments shall
explain not only what the method does, but also its parameters and return value. These are
commented using the @param and @return javadoc tags. The <code> tag shall be used for
Java keywords, names and code samples. This is illustrated in listing 6.1. The @param tag is
used on lines 11 and 23, the @return tag on line 24 and the <code> tag on lines 19, 20 and
23. The html file generated with the javadoc command is, in part, depicted in figure 6.1.

1 /**
2 * Represents an amount of money. Instances are immutable.
3 */
4 public final class Amount {
5 private final int amount;
6
7 /**
8 * Creates a new instance, representing the specified
9 * amount.
10 *
11 * @param amount The amount represented by the newly
12 * created instance.
13 */
14 public Amount(int amount) {
15 this.amount = amount;
16 }
17
18 /**
19 * Subtracts the specified <code>Amount</code> from
20 * this object and returns an <code>Amount</code>
21 * instance with the result.
22 *
23 * @param other The <code>Amount</code> to subtract.
24 * @return The result of the subtraction.
25 */
26 public Amount minus(Amount other) {
27 return new Amount(amount - other.amount);
28 }
29 }

Listing 6.1 Code with javadoc comments.

It is seldom meaningful to add more comments, inside methods. To make sure such com-
ments are up to date is often burdensome extra work, that far too often is simply not done. If
comments are not maintained, they will not correctly describe the code, which will result in
developers not trusting the comments. Low trust in comments is a very unproductive state of
a program. It means both that the commenting work was in vain, and that unnecessary time is
spent reading code instead of comments. Therefore, avoid placing comments inside methods.

82
 Chapter 6 Programming

Figure 6.1 Part of the javadoc generated from the code in listing 6.1.

Instead, the need for comments inside a method should be seen as a signal that the method is
too long, and ought to be split into shorter methods. More on this below.

6.4 Code Smell and Refactoring

The concept code smell describes the state of a particular piece of code. It originates from
[FOW], which, despite its age, is still very relevant. This book describes certain unwanted
states (smells) of a code and how to get rid of them. The way to remove a code smell is to
refactor the code, which means to improve it without changing its functionality. A refactoring
is a well-defined way to change a specific detail of the code, for example to change a method’s
name. [FOW] lists numerous refactorings and tells how to use them to remove different code
smells. This section describes a small number of the many code smells and refactorings men-
tioned in the book. It is of course not necessary to first introduce a smell by making make the
corresponding mistake, and then refactor the code. Better is to learn from the particular smell
and never make the mistake.
The amount of code smell in a program is a quite sure sign of the programmer’s skills.

83
 Chapter 6 Programming

Novice programmers reveal their lack of knowledge by writing code that has several code
smells. It is relatively common for employers to test the ability to find such problems in a
piece of code when hiring new programmers.

Duplicated Code
Identical code in more than one place in the program is a really bad smell. It means whenever
that piece of code shall be changed, exactly the same editing must be done in all locations
where the duplicated code exists. This is of course inefficient since more writing is needed,
but far worse is that it is easy to miss one or more code locations, which means the code will
not work as expected after the (incomplete) change is made. This will lead to long and boring
searches for lines in the program where the duplicated code was not changed as intended.
How sensitive to duplicated code shall one be? The answer is very sensitive! The goal must
always be that not a single statement shall be repeated anywhere in the program. Allowing
duplicated code is to enter a road that leads to disaster. Duplicated code is normally introduced
by copying previously written code. You should hear a loud warning bell ring if you ever type
ctrl-c ctrl-v when programming.
As an example, consider the code in listing 6.2, where the printout of the contents of the
names array is duplicated. In fact, also the javadoc comment to the three methods is dupli-
cated. Duplicated comments introduce exactly the same complications as duplicated code.

1 package se.kth.ict.oodbook.prog.smell;
2
3 /**
4 * This class has bad smell since it contains duplicated code.
5 * The duplicated code is the loop printing the contents of
6 * the <code>names</code> array.
7 */
8 public class ClassWithDuplicatedCode {
9 private String[] names;
10
11 /**
12 * To perform its task, this method has to print the
13 * contents of the <code>names</code> array.
14 */
15 public void aMethodThatShowsNames() {
16 //some code.
17 for (String name : names) {
18 System.out.println(name);
19 }
20 //some code.
21 }
22
23 /**
24 * To perform its task, this method has to print the

84
 Chapter 6 Programming

25 * contents of the <code>names</code> array.

26 */
27 public void otherMethodThatShowsNames() {
28 //some code.
29 for (String name : names) {
30 System.out.println(name);
31 }
32 //some code.
33 }
34
35 /**
36 * To perform its task, this method has to print the
37 * contents of the <code>names</code> array.
38 */
39 public void thirdMethodThatShowsNames() {
40 //some code.
41 for (String name : names) {
42 System.out.println(name);
43 }
44 //some code.
45 }
46 }

Listing 6.2 The loop with the printout of the names array is duplicated.

Suppose the printout in listing 6.2 has to be modified, say that lines 18, 30 and 42 shall
be changed to System.out.println("name: " + name);. This change has to be per-
formed on all three lines. Also, as mentioned above, the fact that there is duplicated code
makes it quite difficult to be sure all lines where the code exists where actually changed,
especially if the program is large.
This smell is removed by using the refactoring Extract Method, which means to move code
from an existing method into a new method, which contains this particular code. In the current
example, it is the printout loop that shall be placed in the newly created method. This new
method is then called on all lines where the printout is required. Listing 6.3 shows the code
after applying this refactoring. Here, there is no duplicated code, the desired change is done
by editing only line 43.

1 package se.kth.ict.oodbook.prog.smell;
2
3 /**
4 * This class does not contain duplicated code. The
5 * previously duplicated code has been extracted to
6 * the method <code>printNames</code>
7 */
8 public class ClassWithoutDuplicatedCode {
9 private String[] names;

85
 Chapter 6 Programming

10
11 /**
12 * To perform its task, this method has to print the
13 * contents of the <code>names</code> array.
14 */
15 public void aMethodThatShowsNames() {
16 //some code.
17 printNames();
18 //some code.
19 }
20
21 /**
22 * To perform its task, this method has to print the
23 * contents of the <code>names</code> array.
24 */
25 public void otherMethodThatShowsNames() {
26 //some code.
27 printNames();
28 //some code.
29 }
30
31 /**
32 * To perform its task, this method has to print the
33 * contents of the <code>names</code> array.
34 */
35 public void thirdMethodThatShowsNames() {
36 //some code.
37 printNames();
38 //some code.
39 }
40
41 private void printNames() {
42 for (String name : names) {
43 System.out.println("name: " + name);
44 }
45 }
46 }

Listing 6.3 When the Extract Method refactoring has been applied, the loop with the
printout of the names array is no longer duplicated.

Listing 6.4 shows a more subtle example of duplicated code. The problem here is the code
sequence[1], which is used to access the first element in the array sequence. This code is
wrong, since the first element is located at index zero, not one. To fix this bug, both lines 16
and 23 must be changed. The situation would be even worse in a larger program, where the
indexing mistake would occur on numerous lines. Just as in the previous example, the solution
is to extract a method containing the duplicated code, see listing 6.5.

86
 Chapter 6 Programming

1 package se.kth.ict.oodbook.prog.smell;
2
3 /**
4 * This class has bad smell since it contains duplicated code.
5 * The duplicated code is <code>sequence[1]</code> to access
6 * the first element in the <code>sequence</code> array.
7 */
8 public class ClassWithUnobviousDuplicatedCode {
9 private int[] sequence;
10
11 /**
12 * @return <code>true</code> if the the specified value is
13 * equal to the first element in the sequence.
14 */
15 public boolean startsWith(int value) {
16 return sequence[1] == value;
17 }
18
19 /**
20 * @return The first element in the sequence array.
21 */
22 public int getFirstElement() {
23 return sequence[1];
24 }
25 }

Listing 6.4 The duplicated code in this class is the usage of sequence[1] to access the first
element in the array

1 package se.kth.ict.oodbook.prog.smell;
2
3 /**
4 * This class does not contain duplicated code. The previously
5 * duplicated code has been extracted to the method
6 * <code>firstElement</code>
7 */
8 public class ClassWithoutUnobviousDuplicatedCode {
9 private int[] sequence;
10
11 /**
12 * @return <code>true</code> if the the specified value is
13 * equal to the first element in the sequence.
14 */
15 public boolean startsWith(int value) {
16 return firstElement() == value;
17 }

87
 Chapter 6 Programming

18
19 /**
20 * @return The first element in the sequence array.
21 */
22 public int getFirstElement() {
23 return firstElement();
24 }
25
26 private int firstElement() {
27 return sequence[0];
28 }
29 }

Listing 6.5 The introduction of the method firstElement has removed the duplicated code.

All occurrences of the duplicated code were located in the same class in both examples
above. This is certainly not always the case, the same code might just as well exist in different
classes. Also in this case, the solution is to extract a method with the duplicated code, and
replace all occurrences of that code with calls to the newly created method. The specific issue
when multiple classes are involved, is where to place the new method. One option is to place
it in one of the classes that had the duplicated code, another option is to place it in a new class.
In either case, the classes that do not contain the new method, must call the new method in the
class where it is placed. The best placement must be decided in each specific case, based on
how cohesion, coupling and encapsulation are affected by the different alternatives.

Long Method
It is easier to understand the code if all methods have names that clearly explain what the
method does. A guideline for deciding if a method is too long is does the method name tell
everything that is needed to fully understand the method body? If there seems to be need for
comments inside a method, that is clearly not the case. In fact, comments or need of comments
inside a method is a clear sign that the method is too long. Thus, what matters is not primarily
the number of lines in a method, but how easy it is to understand the method body. Consider
the method in listing 6.6, which is quite short but still not easy to understand. What is the
meaning of the numbers 65 and 90 on line 13? The answer is that the ASCII numbers of upper
case letters are between 65 and 90. This becomes clear if a new method, with an explaining
name, is introduced, see listing 6.7. To introduce a new method with an explaining name is
almost always the best way to shorten and explain a method that is too long.

1 /**
2 * Counts the number of upper case letters in the specified
3 * string.
4 *
5 * @param source The string in which uppercase letters are
6 * counted.

88
 Chapter 6 Programming

7 * @return The number of uppercase letters in the specified

8 * string.
9 */
10 public int countUpperCaseLetters(String source) {
11 int noOfUpperCaseLetters = 0;
12 for (char letter : source.toCharArray()) {
13 if (letter >= 65 && letter <= 90) {
14 noOfUpperCaseLetters++;
15 }
16 }
17 return noOfUpperCaseLetters;
18 }

Listing 6.6 In spite of the few lines, this method is too long since it is not clear what line 13
does.

1 /**
2 * Counts the number of upper case letters in the specified
3 * string.
4 *
5 * @param source The string in which uppercase letters are
6 * counted.
7 * @return The number of uppercase letters in the specified
8 * string.
9 */
10 public int countUpperCaseLetters(String source) {
11 int noOfUpperCaseLetters = 0;
12 for (char letter : source.toCharArray()) {
13 if (isUpperCaseLetter(letter)) {
14 noOfUpperCaseLetters++;
15 }
16 }
17 return noOfUpperCaseLetters;
18 }
19
20 private boolean isUpperCaseLetter(char letter) {
21 return letter >= 65 && letter <= 90;
22 }

Listing 6.7 Here, each method body is explained by the method’s name.

It is sometimes argued that the program becomes slower if there are many method calls.
This is simply not true, to perform a method call is not significantly slower than any other
statement. Trying to decrease execution time by minimizing the number of method calls is not
any smarter than trying to minimize the number of statements in the program.

89
 Chapter 6 Programming

Large Class
Just as is the case for methods, whether a class is too large is not primarily decided by the
number of lines. The main criteria is instead cohesion, a class is too large if it has bad cohesion.
Cohesion was covered extensively above, in section 5.2. The class in listing 6.8 shows that
cohesion can be improved also by splitting small classes. This listing contains the Meeting
class, which represents a meeting in a calendar. It has the fields startTime and endTime
that together define the meeting’s duration. These two fields are more closely related to each
other, than to other fields in the class. The fact that they have a common suffix, Time, helps us
see this. Cohesion is improved in listing 6.9, by extracting a class with these two fields. This
is a quite common way to realize that a new class is appropriate. As programming continues,
the new class will probably get more fields and methods.

1 package se.kth.ict.oodbook.prog.smell;
2
3 import java.time.LocalDateTime;
4
5 /**
6 * This class represents a meeting in a calendar.
7 */
8 public class MeetingLowerCohesion {
9 private LocalDateTime startTime;
10 private LocalDateTime endTime;
11 private String name;
12 private boolean alarmIsSet;
13
14 //More fields and methods.
15 }

Listing 6.8 This class has two fields that are more closely related than other
fields. This is an indication that cohesion can be improved by extracting a
new class, with these fields.

1 package se.kth.ict.oodbook.prog.smell;
2
3 /**
4 * This class represents a meeting in a calendar.
5 */
6 public class MeetingHigherCohesion {
7 private TimePeriod period;
8 private String name;
9 private boolean alarmIsSet;
10
11 //More fields and methods.
12 }
13

90
 Chapter 6 Programming

14
15 package se.kth.ict.oodbook.prog.smell;
16
17 import java.time.LocalDateTime;
18
19 /**
20 * Represents a period in time, with specific start
21 * and end time.
22 */
23 class TimePeriod {
24 private LocalDateTime startTime;
25 private LocalDateTime endTime;
26
27 //More fields and methods.
28 }

Listing 6.9 Here, cohesion is improved by moving the related fields to a new
class.

Long Parameter List

Long parameter lists are hard to understand, because it is difficult to remember the meaning of
each parameter, especially if there are many parameters of the same type. It is also very likely
that a long parameter list changes, and changed parameters means changed public interface.
The need to change the list often arises because it is long since it consists of primitive data, not
objects. This means there is no encapsulation in the list, whenever the need of data changes in
the method, it is reflected in its parameter list.
This bad smell can often be removed with the refactorings Preserve Whole Object or In-
troduce Parameter Object, which both replace primitive data with objects. The former is
illustrated in listing 6.10, that has a long parameter list (line 22), and listing 6.11, where the
parameter list is shortened by passing an entire object instead of the fields in that object (line
21). The latter is illustrated in listing 6.12, which has a long parameter list (line 16), and list-
ing 6.13, where the list is shortened by introducing a new object that encapsulates parameters
(line 15). It is quite common to discover that this newly created class was needed, and that
either existing or new methods belong there.

1 package se.kth.ict.oodbook.prog.smell;
2
3 /**
4 * This class represents a person. The call to
5 * <code>dbHandler</code> does not preserve the
6 * <code>Person</code> object. The fields are instead passed as
7 * primitive parameters.
8 */
9 public class PersonObjectNotPreserved {

91
 Chapter 6 Programming

10 private String name;

11 private String address;
12 private String phone;
13
14 /**
15 * Saves this <code>Person</code> to the specified
16 * database.
17 *
18 * @param dbHandler The database handler used to save the
19 * <code>Person</code>.
20 */
21 public void savePerson(DBHandler dbHandler) {
22 dbHandler.savePerson(name, address, phone);
23 }
24 }

Listing 6.10 The call to dbHandler does not preserve the Person object. The fields are instead
passed as primitive parameters.

1 package se.kth.ict.oodbook.prog.smell;
2
3 /**
4 * This class represents a person. The call
5 * to <code>dbHandler</code> preserves the
6 * <code>Person</code> object.
7 */
8 public class PersonObjectPreserved {
9 private String name;
10 private String address;
11 private String phone;
12
13 /**
14 * Saves this <code>Person</code> to the specified
15 * database.
16 *
17 * @param dbHandler The database handler used to save the
18 * <code>Person</code>.
19 */
20 public void savePerson(DBHandler dbHandler) {
21 dbHandler.savePerson(this);
22 }
23 }

Listing 6.11 The call to dbHandler preserves the Person object.

1 package se.kth.ict.oodbook.prog.smell;

92
 Chapter 6 Programming

2
3 /**
4 * This class represents a bank account. The
5 * <code>deposit</code> method takes primitive parameters
6 * instead of using a parameter object.
7 */
8 public class AccountWithoutParameterObject {
9 /**
10 * Adds the specified amount of the specified currency to
11 * the balance.
12 *
13 * @param currency The currency of the deposited amount.
14 * @param amount The amount to deposit.
15 */
16 public void deposit(String currency, int amount) {
17 }
18 }

Listing 6.12 The deposit method takes primitive parameters instead of using a parameter ob-
ject.

1 package se.kth.ict.oodbook.prog.smell;
2
3 /**
4 * This class represents a bank account. The parameters of the
5 * <code>deposit</code> method are encapsulated in an object.
6 */
7 public class AccountWithParameterObject {
8 /**
9 * Adds the specified amount of the specified currency to
10 * the balance.
11 *
12 * @param currency The currency of the deposited amount.
13 * @param amount The amount to deposit.
14 */
15 public void deposit(Amount amount) {
16 }
17 }
18
19 package se.kth.ict.oodbook.prog.smell;
20
21 /**
22 * Represents an amount
23 */
24 class Amount {
25 private String currency;

93
 Chapter 6 Programming

26 private int amount;

27
28 //Methods
29 }

Listing 6.13 The Amount class has been created and encapsulates the parameters of the deposit
method.

In some cases, there are parameters that are simply not needed, because the called method
itself can find the data by making a request to an object it already knows. This refactoring
is called Replace Parameter With Method, and is illustrated in listing 6.14, which has the
unnecessary parameter (lines 19 and 39), and listing 6.15, where the called method gets the
data instead of using a parameter (lines 18 and 39).

1 package se.kth.ict.oodbook.prog.smell;
2
3 /**
4 * The bank application’s controller. The call to
5 * <code>withdraw</code> passes the <code>fee</code> parameter
6 * that is not needed.
7 */
8 public class ControllerPassingExtraParameter {
9 private AccountWithExtraParameter account;
10 private AccountCatalog accts;
11
12 /**
13 * Withdraws the specified amount.
14 *
15 * @param amount The amount to withdraw.
16 */
17 public void withdraw(Amount amount) {
18 Amount fee = accts.getWithDrawalFeeOfAccount(account);
19 account.withdraw(amount, fee);
20 }
21 }
22
23 package se.kth.ict.oodbook.prog.smell;
24
25 /**
26 * Represents a bank account. The method <code>withdraw</code>
27 * takes the <code>fee</code> parameter that is not needed.
28 */
29 public class AccountWithExtraParameter {
30 // Needed for some unknown purpose.
31 private AccountCatalog acctSpecs;
32
33 /**

94
 Chapter 6 Programming

34 * Withdraws the specified amount.

35 *
36 * @param amount The amount to withdraw.
37 * @param fee The withdrawal cost.
38 */
39 public void withdraw(Amount amount, Amount fee) {
40 }
41 }

Listing 6.14 The call to withdraw passes the fee parameter that is not needed.

1 package se.kth.ict.oodbook.prog.smell;
2
3 /**
4 * The bank application’s controller. The call to
5 * <code>withdraw</code> does not pass the
6 * <code>fee</code> parameter.
7 */
8 public class ControllerNotPassingExtraParameter {
9 private AccountWithoutExtraParameter account;
10 private AccountCatalog accts;
11
12 /**
13 * Withdraws the specified amount.
14 *
15 * @param amount The amount to withdraw.
16 */
17 public void withdraw(Amount amount) {
18 account.withdraw(amount);
19 }
20 }
21
22 package se.kth.ict.oodbook.prog.smell;
23
24 /**
25 * Represents a bank account. The <code>fee</code> parameter
26 * is not passed to <code>withdraw</code>, since it can be
27 * retrieved in that method itself.
28 */
29 public class AccountWithoutExtraParameter {
30 // Needed for some unknown purpose, besides getting the
31 // withdrawal fee.
32 private AccountCatalog acctSpecs;
33
34 /**
35 * Withdraws the specified amount.

95
 Chapter 6 Programming

36 *
37 * @param amount The amount to withdraw.
38 */
39 public void withdraw(Amount amount) {
40 acctSpecs.getWithDrawalFeeOfAccount(this);
41 }
42 }

Listing 6.15 The call to withdraw does not pass the fee parameter, since it is retrieved by the
withdraw method.

Excessive Use of Primitive Variables

This code smell is called Primitive Obsession in [FOW]. Many advantages of using objects
instead of primitive data have already been mentioned. Primitive data is not forbidden, but
should be used with care. Excessive primitive data means that everything related to object
oriented development is just thrown in the wastebin. There is no encapsulation at all, no
cohesion, no possibility to minimize coupling, etc.
A long list of fields in a class is a sign that there are too few classes in the program. The
class with the many fields probably has low cohesion, and some of the fields fit better together,
in a new class. This refactoring, Extract Class, is illustrated in listings 6.16, where there are
many fields on lines 8-13, and 6.17, where some of the fields are encapsulated in a new class.

1 package se.kth.ict.oodbook.prog.smell;
2
3 /**
4 * Represents a person. This class has excessive primitive
5 * data, since it has fields that fit better as an object.
6 */
7 public class PersonManyFields {
8 private String name;
9 private String street;
10 private int zip;
11 private String city;
12 private String phone;
13 private String email;
14
15 // More code in the class.
16 }

Listing 6.16 This class has excessive primitive data, since it has fields that fit better as an object.

1 package se.kth.ict.oodbook.prog.smell;
2
3 /**

96
 Chapter 6 Programming

4 * Represents a person. This class uses objects for the fields.

5 */
6 public class PersonFewerFields {
7 private String name;
8 private Address address;
9 private String phone;
10 private String email;
11
12 // More code in the class.
13 }
14
15 package se.kth.ict.oodbook.prog.smell;
16
17 /**
18 * An address where a person lives.
19 */
20 class Address {
21 private String street;
22 private int zip;
23 private String city;
24
25 // More code in the class.
26 }

Listing 6.17 This Person class uses objects for the fields.

It might be that a class has not only fields, but field(s) and one or more methods that are
closer related than other fields and methods in the class. Also in this case, cohesion can be
improved by introducing a new class. The new class shall contain the field(s) and methods of
the original class that belong closely together. This refactoring is called Replace Data Value
With Object. The code before the applying the refactoring is listed in listing 6.18. It shows the
class Person, which has the pnr field (line 10) that holds a person number. The class also
has the method validatePnr (line 17), that checks if the control digit of the person number
is correct. This method really belongs to the pnr field, not to the Person class. Cohesion is
improved in listing 6.19, by introducing the PersonNumber class, which shall always be used
to represent person numbers. Note that validatePnr (line 30) is called in the constructor of
PersonNumber (line 26). That way, there can never exist any invalid person numbers in the
program, they are immediately revealed when a PersonNumber is created.

1 package se.kth.ict.oodbook.prog.smell;
2
3 /**
4 * Represents a person. This class has low cohesion since the
5 * method <code>validatePnr</code> belongs to the
6 * <code>pnr</code> field, rather than to this class.
7 */

97
 Chapter 6 Programming

8 public class PersonWithoutPnrClass {

9 private String name;
10 private String pnr;
11
12 public PersonWithoutPnrClass(String name, String pnr) {
13 this.name = name;
14 this.pnr = pnr;
15 }
16
17 private void validatePnr(String pnr) {
18 }
19 }

Listing 6.18 This class has low cohesion since the method validatePnr belongs to the pnr
field, rather than to this class.

1 package se.kth.ict.oodbook.prog.smell;
2
3 /**
4 * Represents a person. The method <code>validatePnr</code>
5 * has been moved to <code>PersonNumber</code>.
6 */
7 public class PersonWithPnrClass {
8 private String name;
9 private PersonNumber pnr;
10
11 public PersonWithPnrClass(String name, PersonNumber pnr) {
12 this.name = name;
13 this.pnr = pnr;
14 }
15 }
16
17 package se.kth.ict.oodbook.prog.smell;
18
19 /**
20 * Represents a person number.
21 */
22 public class PersonNumber {
23 private String pnr;
24
25 public PersonNumber(String pnr) {
26 validatePnr(pnr);
27 this.pnr = pnr;
28 }
29
30 private void validatePnr(String pnr) {

98
 Chapter 6 Programming

31 }
32 }

Listing 6.19 The method validatePnr has been moved to PersonNumber.

A far too common mistake is to use an array of primitive data instead of an object. If arrays
are used correctly, all elements in the same array means the same thing. It is not correct if
different elements in an array means different things, as is the case with the stats array in
listing 6.20. In this code, there is absolutely nothing showing that the first element is the name
of a football team, the second the number of wins, the third the number of draws and the fourth
the number of losses. This information exists only in the mind of the developer, and it is of
course easy to confuse the meaning of the array positions. The code has been improved in
listing 6.21, where an object is used instead of the array. The meaning of each value is now
clear from the names of the fields in the object.

1 String[] stats = new String[3];

2 stats[0] = "Hammarby";
3 stats[1] = "2";
4 stats[2] = "1";
5 stats[3] = "2";
6 printStatistics(stats);

Listing 6.20 This code smells, because

of the array stats where different po-
sitions have different meaning.

1 Stats stats = new Stats("Hammarby", 2, 1, 2);

2 printStatistics(stats);
3
4 public class Stats {
5 private String name;
6 private int wins;
7 private int draws;
8 private int losses;
9
10 public Stats(String name, int wins, int draws,
11 int losses) {
12 this.name = name;
13 this.wins = wins;
14 this.draws = draws;
15 this.losses = losses;
16 }
17 }

Listing 6.21 This code encapsulates the values in an object.

99
 Chapter 6 Programming

Many programming languages, including Java, has enumerations. This enables defining a
custom type and the possible values of that type. As an example, consider listing 6.22 that
does not use an enumerator. Instead, the possible results of the call to connect are strings.
With such code, nasty bugs can appear because of misspelling the result string. An equally
bad alternative is to use integers for the outcomes of connect. In this case, bugs might appear
bescause of confusing which number means what. A much better alternative is to introduce a
new type for the outcomes, using an enumeration. This is illustrated in listing 6.23. The new
type, ResultCode, can take the values SUCCESS, PENDING and FAILURE. The meaning of
each outcome is obvious from the value and misspelled values will generate a compiler error.

1 String result = connect();

2
3 if (result.equals("SUCCESS")) {
4 // Handle established connection.
5 } else if (result.equals("PENDING")) {
6 // Handle pending connection.
7 } else if (result.equals("FAILURE")) {
8 // Handle connection failure.
9 } else {
10 // Something went wrong.
11 }

Listing 6.22 Using strings to represent constants.

1 Outcome result = connect();

2
3 if (result == Outcome.SUCCESS) {
4 // Handle established connection.
5 } else if (result == Outcome.PENDING) {
6 // Handle pending connection.
7 } else if (result == Outcome.FAILURE) {
8 // Handle connection failure.
9 } else {
10 // Something went wrong.
11 }
12
13 /**
14 * Defines possible outcomes of a connection attempt.
15 */
16 public enum Outcome {
17 SUCCESS, PENDING, FAILURE
18 }

Listing 6.23 Using an enum to represent constants.

100
 Chapter 6 Programming

Meaningless Names
This is not mentioned as a particular smell in [FOW], but should still be avoided at all costs.
Everything that is declared in a program (packages, classes, interfaces, methods, fields, pa-
rameters, local variables, etc) must have a meaningful name. This can not be stressed enough.
The following list provides a few naming guidelines.

• Do not use one-letter identifiers like Person p = new Person();, instead write
Person person = Person();. There are two exceptions to this guideline. The
first is when the full name of the abstraction is just one letter long, it is for example
appropriate to use the identifier x for an x coordinate. The second exception is that a
one letter identifier is accepted for a loop variable, which is normally named i. Nested
loops are named using following letters in alphabetical order, j, k, etc.

• Do not name a temporarily used variable tmp or temp (unless it represents a tempera-
ture). For example do not swap two values as in listing 6.24, instead use variable names
like in listing 6.25.

• Never distinguish similar identifiers by appending numbers. As an example, when trans-

ferring money between two bank accounts, they could be named fromAccount and
toAccount, but not account1 and account2.

• Do not be afraid of long names, what matters is that the identifier correctly explain
the purpose of what is named. Say for example that some reward is given to the first
customer buying a particular item in some campaign in a shop. An adequate name for a
variable holding that customer could be firstCustomerBuyingCampaignItem.

1 int tmp = varsToSwap[0];

2 varsToSwap[0] = varsToSwap[1];
3 varsToSwap[1] = tmp;

Listing 6.24 A variable is erronously

named tmp, just because it is used
temporarily.

1 int valAtIndexZero = varsToSwap[0];

2 varsToSwap[0] = varsToSwap[1];
3 varsToSwap[1] = valAtIndexZero;

Listing 6.25 The temporary variable has, cor-

rectly, a name describing its purpose.

101
 Chapter 6 Programming

Unnamed Values
This is not mentioned as a particular smell in [FOW], but should still be avoided at all costs.
All values in a program shall have an explaining name. Never introduce a value in a statement
without naming it first, even if that statement is the only place the value is used. Without a
name it can be very hard to understand the purpose of the value. Naming the value is a better
practice than writing a comment to explain it. A name is part of the program, the compiler
helps to ensure that the name is used correctly. A comment, on the other hand, is a kind of
duplicated information that exists besides the program. There is always the risk that comments
are not maintained when the program changes. Below is a list with some examples of naming
values.

• Say that the method connect(int timeout) tries to connect for timeout number
of milliseconds before stopping. Say also that we want to make it try for ten sec-
onds. A straightforward way to write this could be connect(10000), but then the
reader gets no information about the purpose of the value 10000. A better way is to
write connect(10 * MILLIS_PER_SECOND), and defining the constant private
static final int MILLIS_PER_SECOND = 1000;. Still, however, the purpose
of the value 10 might not be clear. The best way to code this is to also define a constant
private static final int CONNECT_TIMEOUT_SECS = 10;, or, if it is not a
constant, the variable int connectTimeoutSecs = 10;. Now, the code becomes
connect(CONNECT_TIMEOUT_SECS * MILLIS_PER_SECOND) or
connect(connectTimeoutSecs * MILLIS_PER_SECOND).

• The practice of naming values applies (at least) to all primitive types, and also to strings,
since a string can be written as a primitive value, without using the keyword new. Con-
sider for example opening a file, whose name is in the variable fileName, located in
the directory whose name is in the variable dirName. Assuming that there is a method
openFile, that opens a file, this might be done with the statement openFile(dirName
+ "\" + fileName);. The meaning of the value "\" might seem clear, still, it is
even clearer to introduce the constant private static final String
PATH_SEPARATOR = "\";, and write openFile(dirName + PATH_SEPARATOR
+ fileName);. Using this constant everywhere a path separator is needed also gives
the advantage that it is easy to change path separator, if running on a system where the
path separator is not a backslash. Using the constant, only one line has to be changed,
the declaration of the constant.

• It is not required to use a variable or constant to name the value, sometimes a method
suits better. This is often the case when naming values that occur in if statements. As
an example, consider an if statement checking for end of line (EOL) in a string. EOL
in a unix file is represented by a character with ASCII code 10, therefore, the code in
listing 6.26 might be used. This code is unclear, why check for the value 10? A better
solution is to introduce the method isUnixEol, and use the code in figure 6.27.

102
 Chapter 6 Programming

1 /**
2 * Finds the index of the first Unix EOL in the specified
3 * string.
4 *
5 * @param source The string in which to look for EOL.
6 * @return The index of the first EOL, or -1 if there was
7 * no EOL in the specified string.
8 */
9 public int findIndexOfFirstEolWorse(String source) {
10 char[] sourceChars = source.toCharArray();
11 for (int i = 0; i < sourceChars.length; i++) {
12 if (sourceChars[i] == 10) {
13 return i;
14 }
15 }
16 return -1;
17 }

Listing 6.26 It is quite difficult to understand the meaning of an unnamed value, like the
value 10 on line 12

1 private boolean isUnixEol(char character) {

2 return character == 10;
3 }
4
5 /**
6 * Finds the index of the first Unix EOL in the specified
7 * string.
8 *
9 * @param source The string in which to look for EOL.
10 * @return The index of the first EOL, or -1 if there was
11 * no EOL in the specified string.
12 */
13 public int findIndexOfFirstEolBetter(String source) {
14 char[] sourceChars = source.toCharArray();
15 for (int i = 0; i < sourceChars.length; i++) {
16 if (isUnixEol(sourceChars[i])) {
17 return i;
18 }
19 }
20 return -1;
21 }

Listing 6.27 The purpose of the value 10 on line 2 is explained by the name of the method,
isUnixEol

103
 Chapter 6 Programming

6.5 Coding Case Study

Now, it is finally time to write the RentCar program. This section does not include a complete
listing of the entire program. That can be found in the accompanying NetBeans project, which
can be downloaded from GitHub [Code]. Here follows a description of the first parts of the
code, parts where particular afterthought was needed, and where the design was not followed.
Even though this is a quite small program, the design is still big enough to make it difficult
to decide where to start coding. This is a result of having designed the entire requirements
specification in one go, without any intermediate coding. Normally, each system operation
would have been coded as soon as the design was finished. To implement a design in code
is the only way to get a full understanding of its strengths and weaknesses. The best way
to implement the design is still to code one system operation at a time, in the order they are
executed. That makes it possible to test run each part of the program as soon as it is written.
However, there is of course no reason to reiterate all the considerations made during design.
Therefore, the final version of each system operation is implemented.

The searchMatchingCar system operation

The first system operation is searchMatchingCar, and the final design is in figure 5.27.
The final version of the start sequence, however, is not in figure 5.28. It was changed in
figure 5.35, where the class RegistryCreator was added. Most of the coding is a very
straightforward implementation of those diagrams, listed in figures below, but three things
require extra attention. First, nothing has been decided on the implementation of the car
registry. Where are the cars stored? The solution is to use a list with some hard coded cars,
see lines 11, 14 and 46-53 in figure 6.31. This is enough for testing purposes. Second, neither
requirements specification, nor design, are very specific on the search algorithm when looking
for a matching car. To which extent must the search criteria be met to consider a car to
match them? Can search criteria be ignored or must they all be specified? This should of
course be discussed with the customer. The chosen implementation, se lines 49-69 in figure
6.30, requires exact match of all specified parameters, but gives the possibility to leave all
parameters except AC and four wheel drive unspecified. Third, and last, what shall happen
when the program is executed? Since there is no view, to get some output and be able to verify
the functionality, the method sampleExecution is added to View, as can be seen on lines
27-44 in figure 6.28. This method contains hard coded calls to all system operations and prints
the result of those calls. Also, a toString method has been added to CarDTO to provide an
informative printout of objects of that class, see lines 71-81 in figure 6.30.

1 package se.kth.ict.rentcar.view;
2
3 import se.kth.ict.rentcar.controller.Controller;
4 import se.kth.ict.rentcar.integration.CarDTO;
5
6 /**
7 * This program has no view, instead, this class is a

104
 Chapter 6 Programming

8 * placeholder for the entire view.

9 */
10
11 public class View {
12 private Controller contr;
13
14 /**
15 * Creates a new instance.
16 *
17 * @param contr The controller that is used for all
18 * operations.
19 */
20 public View(Controller contr) {
21 this.contr = contr;
22 }
23
24 /**
25 * Simulates a user input that generates calls to all
26 * system operations.
27 */
28 public void sampleExecution() {
29 CarDTO unavailableCar =
30 new CarDTO(1000, "nonExistingSize", true, true,
31 "red", null);
32 CarDTO availableCar =
33 new CarDTO(1000, "medium", true, true, "red",
34 null);
35
36 CarDTO foundCar =
37 contr.searchMatchingCar(unavailableCar);
38 System.out.println(
39 "Result of searching for unavailable car: " +
40 foundCar);
41 foundCar = contr.searchMatchingCar(availableCar);
42 System.out.println(
43 "Result of searching for available car: " +
44 foundCar);
45 }
46 }

Listing 6.28 The class View when only the searchMatchingCar system operation has
been implemented.

1 package se.kth.ict.rentcar.controller;
2
3 import se.kth.ict.rentcar.integration.CarRegistry;

105
 Chapter 6 Programming

4 import se.kth.ict.rentcar.integration.CarDTO;
5 import se.kth.ict.rentcar.integration.RegistryCreator;
6
7 /**
8 * This is the application’s only controller class. All
9 * calls to the model pass through here.
10 */
11
12 public class Controller {
13 private CarRegistry carRegistry;
14
15 /**
16 * Creates a new instance.
17 *
18 * @param regCreator Used to get all classes that
19 * handle database calls.
20 */
21 public Controller(RegistryCreator regCreator) {
22 this.carRegistry = regCreator.getCarRegistry();
23 }
24
25 /**
26 * Search for a car matching the specified search criteria.
27 *
28 * @param searchedCar This object contains the search
29 * criteria. Fields in the object that
30 * are set to <code>null</code> or
31 * <code>0</code> are ignored.
32 * @return The best match of the search criteria.
33 */
34 public CarDTO searchMatchingCar(CarDTO searchedCar) {
35 return carRegistry.findCar(searchedCar);
36 }
37 }

Listing 6.29 The class Controller when only the searchMatchingCar system operation has
been implemented.

1 package se.kth.ict.rentcar.integration;
2 /**
3 * Contains information about one particular car.
4 */
5 public final class CarDTO {
6 private final int price;
7 private final String size;
8 private final boolean AC;

106
 Chapter 6 Programming

9 private final boolean fourWD;

10 private final String color;
11 private final String regNo;
12
13 /**
14 * Creates a new instance representing a particular car.
15 *
16 * @param price The price paid to rent the car.
17 * @param size The size of the car, e.g.,
18 * <code>medium</code>.
19 * @param AC <code>true</code> if the car has air
20 * condition.
21 * @param fourWD <code>true</code> if the car has four
22 * wheel drive.
23 * @param color The color of the car.
24 * @param regNo The car’s registration number.
25 */
26 public CarDTO(int price, String size, boolean AC,
27 boolean fourWD, String color, String regNo) {
28 this.price = price;
29 this.size = size;
30 this.AC = AC;
31 this.fourWD = fourWD;
32 this.color = color;
33 this.regNo = regNo;
34 }
35
36 /**
37 * Checks if the specified car has the same features as
38 * this car. Fields that are set to <code>null</code> or
39 * <code>0</code> are ignored. Note that the check is
40 * for matching features, not for identity. Therefore,
41 * registration number is ignored.
42 *
43 * @param searched Contains search criteria.
44 * @return <code>true</code> if this object has the same
45 * features as <code>searched</code>,
46 * <code>false</code> if it has not.
47 */
48 boolean matches(CarDTO searched) {
49 if (searched.getPrice() != 0 &&
50 searched.getPrice() != price) {
51 return false;
52 }
53 if (searched.getSize() != null &&
54 !searched.getSize().equals(size)) {

107
 Chapter 6 Programming

55 return false;
56 }
57 if (searched.getColor() != null &&
58 !searched.getColor().equals(color)) {
59 return false;
60 }
61 if (searched.isAC() != AC) {
62 return false;
63 }
64 if (searched.isFourWD() != fourWD) {
65 return false;
66 }
67 return true;
68 }
69
70 @Override
71 public String toString() {
72 StringBuilder builder = new StringBuilder();
73 builder.append("regNo: " + regNo + ", ");
74 builder.append("size: " + size + ", ");
75 builder.append("price: " + price + ", ");
76 builder.append("AC: " + AC + ", ");
77 builder.append("4WD: " + fourWD + ", ");
78 builder.append("color: " + color);
79 return builder.toString();
80 }
81
82 // Getters are not listed.
83
84 }

Listing 6.30 The class CarDTO when only the searchMatchingCar system operation has been
implemented.

1 package se.kth.ict.rentcar.integration;
2
3 import java.util.ArrayList;
4 import java.util.List;
5
6 /**
7 * Contains all calls to the data store with cars that may be
8 * rented.
9 */
10 public class CarRegistry {
11 private List<CarDTO> cars = new ArrayList<>();
12

108
 Chapter 6 Programming

13 CarRegistry() {
14 addCars();
15 }
16
17 /**
18 * Search for a car matching the specified search criteria.
19 *
20 * @param searchedCar This object contains the search
21 * criteria. Fields in the object that
22 are set to <code>null</code> or
23 * <code>0</code> are ignored.
24 * @return <code>true</code> if a car with the same
25 * features as <code>searchedCar</code> was found,
26 * <code>false</code> if no such car was found.
27 */
28 public CarDTO findCar(CarDTO searchedCar) {
29 for (CarDTO car : cars) {
30 if (car.matches(searchedCar)) {
31 return car;
32 }
33 }
34 return null;
35 }
36
37 private void addCars() {
38 cars.add(new CarDTO(1000, "medium", true, true, "red",
39 "abc123"));
40 cars.add(new CarDTO(2000, "large", false, true, "blue",
41 "abc124"));
42 cars.add(new CarDTO(500, "medium", false, false, "red",
43 "abc125"));
44 }
45 }

Listing 6.31 The class CarRegistry when only the searchMatchingCar system operation
has been implemented.

1 package se.kth.ict.rentcar.integration;
2 /**
3 * This class is responsible for instantiating all registries.
4 */
5 public class RegistryCreator {
6 private CarRegistry carRegistry = new CarRegistry();
7
8 /**
9 * Get the value of carRegistry

109
 Chapter 6 Programming

10 *
11 * @return the value of carRegistry
12 */
13 public CarRegistry getCarRegistry() {
14 return carRegistry;
15 }
16 }
Listing 6.32 The class RegistryCreator when only the searchMatchingCar system op-
eration has been implemented.

1 package se.kth.ict.rentcar.startup;
2
3 import se.kth.ict.rentcar.controller.Controller;
4 import se.kth.ict.rentcar.integration.RegistryCreator;
5 import se.kth.ict.rentcar.view.View;
6
7 /**
8 * Contains the <code>main</code> method. Performs all startup
9 * of the application.
10 */
11 public class Main {
12 /**
13 * Starts the application.
14 *
15 * @param args The application does not take any command
16 * line parameters.
17 */
18 public static void main(String[] args) {
19 RegistryCreator creator = new RegistryCreator();
20 Controller contr = new Controller(creator);
21 new View(contr).sampleExecution();
22 }
23 }
Listing 6.33 The class Main when only the searchMatchingCar system operation has been
implemented.

The registerCustomer system operation

The code for the registerCustomer system operation is a very straightforward implemen-
tation of figure 5.31. There is only one thing that requires a bit of consideration, namely what
to do with the DTOs in the Rental constructor. For now, there is no reason to do anything
at all, except to save them as fields in the constructed Rental object. Since all DTOs are
immutable, there is no risk that any other object changes the content of a DTO. Remember
that being immutable means an object can not change state, for example because the class

110
 Chapter 6 Programming

itself and all its fields are final. If DTOs had not been final, it would have been suicide to just
keep them in Rental. In that case, the object that sent the DTO to Rental could have kept a
reference to the same DTO object, and later updated it.
There is also another issue with keeping the DTOs. Is it really sure they are just DTOs,
having no logic at all? If, for example, there is the need to validate the driving license number,
or to calculate a person’s age based on the driving license number, the methods performing this
would, with the argument of cohesion, be placed in DrivingLicenseDTO or CustomerDTO.
This would turn that object into an entity object, with business logic, instead of a DTO. This
change is not just a matter of renaming, e.g., from CustomerDTO to Customer, but also
concerns how the object is handled. A DTO may be used in the view, but an entity object may
not. To conclude this discussion, it is obvious that att objects named DTO are, at least for the
moment, DTOs. Therefore, it is perfectly safe to leave them like that now, but we must be
aware that this might have to be changed in the future.
Rental is listed in listing 6.34, to illustrate the reasoning above. The rest of the register
Customer implementation can be found in the accompanying NetBeans project [Code].

1 package se.kth.ict.rentcar.model;
2
3 /**
4 * Represents one particular rental transaction, where one
5 * particular car is rented by one particular customer.
6 */
7 public class Rental {
8 private CustomerDTO customer;
9
10 /**
11 * Creates a new instance, representing a rental made by
12 * the specified customer.
13 *
14 * @param customer The renting customer.
15 */
16 public Rental(CustomerDTO customer) {
17 this.customer = customer;
18 }
19 }

Listing 6.34 The class Rental, after implementing the registerCustomer system opera-
tion.

The bookCar system operation

The bookCar design can be found in figures 5.33 and 5.35. Implementing this system opera-
tion reveals one serious problem, the implementation of the setBookedStateOfCar method
in CarRegistry. The cars in the registry are currently stored in a list of CarDTOs. How to
mark that one of them is booked, and not available for rental? This problem reveals a flaw in

111
 Chapter 6 Programming

the implementation of CarRegistry. That class is supposed to call a database or some other
system that stores car data persistently. Such a datastore does not hold a list of immutable
DTOs, but instead raw, mutable data. This data shall not be object-oriented, since it mimics
a store with primitive data. Instead of having methods, objects shall have only primitive vari-
ables. Therefore, the list in CarRegistry is changed, to hold objects of a class CarData,
which has just primitive fields, no methods at all. This class shall not be used anywhere out-
side CarRegistry, since it mimics the contents of the CarRegistry datastore. To ensure it
is not used anywhere else, it is a private inner class, see lines 97-117 in listing 6.35.

1 package se.kth.ict.rentcar.integration;
2
3 import java.util.ArrayList;
4 import java.util.List;
5
6 /**
7 * Contains all calls to the data store with cars that may be
8 * rented.
9 */
10 public class CarRegistry {
11 private List<CarData> cars = new ArrayList<>();
12
13 CarRegistry() {
14 addCars();
15 }
16
17 /**
18 * Search for a car matching the specified search criteria.
19 *
20 * @param searchedCar This object contains the search
21 * criteria. Fields in the object
22 * that are set to <code>null</code>
23 * or <code>0</code> are ignored.
24 * @return <code>true</code> if a car with the same
25 * features as <code>searchedCar</code> was found,
26 * <code>false</code> if no such car was found.
27 */
28 public CarDTO findAvailableCar(CarDTO searchedCar) {
29 for (CarData car : cars) {
30 if (matches(car, searchedCar) && !car.booked) {
31 return new CarDTO(car.regNo, car.price,
32 car.size, car.AC,
33 car.fourWD, car.color);
34 }
35 }
36 return null;
37 }

112
 Chapter 6 Programming

38
39 /**
40 * If there is an existing car with the registration
41 * number of the specified car, set its booked
42 * property to the specified value. Nothing is changed
43 * if the car’s booked property already had the specified
44 * value.
45 *
46 * @param car The car that shall be marked as
47 * booked.
48 * @param bookedState The new value of the booked property.
49 */
50 public void setBookedStateOfCar(CarDTO car,
51 boolean bookedState) {
52 CarData carToBook = findCarByRegNo(car);
53 carToBook.booked = bookedState;
54 }
55
56 private void addCars() {
57 cars.add(new CarData("abc123", 1000, "medium", true,
58 true, "red"));
59 cars.add(new CarData("abc124", 2000, "large", false,
60 true, "blue"));
61 cars.add(new CarData("abc125", 500, "medium", false,
62 false, "red"));
63 }
64
65 private boolean matches(CarData found, CarDTO searched) {
66 if (searched.getPrice() != 0 &&
67 searched.getPrice() != found.price) {
68 return false;
69 }
70 if (searched.getSize() != null &&
71 !searched.getSize().equals(found.size)) {
72 return false;
73 }
74 if (searched.getColor() != null &&
75 !searched.getColor().equals(
76 found.color)) {
77 return false;
78 }
79 if (searched.isAC() != found.AC) {
80 return false;
81 }
82 if (searched.isFourWD() != found.fourWD) {
83 return false;

113
 Chapter 6 Programming

84 }
85 return true;
86 }
87
88 private CarData findCarByRegNo(CarDTO searchedCar) {
89 for (CarData car : cars) {
90 if (car.regNo.equals(searchedCar.getRegNo())) {
91 return car;
92 }
93 }
94 return null;
95 }
96
97 private static class CarData {
98 private String regNo;
99 private int price;
100 private String size;
101 private boolean AC;
102 private boolean fourWD;
103 private String color;
104 private boolean booked;
105
106 public CarData(String regNo, int price, String size,
107 boolean AC, boolean fourWD,
108 String color) {
109 this.regNo = regNo;
110 this.price = price;
111 this.size = size;
112 this.AC = AC;
113 this.fourWD = fourWD;
114 this.color = color;
115 this.booked = false;
116 }
117 }
118 }

Listing 6.35 The class CarRegistry, after implementing the bookCar system operation.

With the above change to CarRegistry, it becomes necessary to change the CarDTO
method matches, which compares the features the customer wishes with the features of an
available car. It must now compare fields in a CarDTO with fields in a CarData, and the latter
must not be used outside CarRegistry. This is solved by removing matches from CarDTO
and instead making it a private method in CarRegistry, see lines 65-86 in listing 6.35. This
is in fact a better location for matches. First, the total public interface decreases since a pub-
lic method becomes private. Second, it was never a very good idea to have such a method in
a DTO. A DTO shall not have any business logic, and matches can be regarded as business

114
 Chapter 6 Programming

logic, it performs a matching algorithm and is not just a simple data comparison.
Since a booked car can not be rented by other customers, the findCar method must not
return a booked car, even if it matches the search criteria. This results in the if-statement
on line 30 in listing 6.35. To clarify this new behavior, the method name is changed to
findAvailableCar.
Another refactoring was to change the order of the parameters in the CarDTO constructor,
regNo is now the first parameter. This was done because every time that constructor was
called, the first thought was to place regNo first. This is a clear sign that this is a more logical
ordering of the parameters.
Also the Rental constructor had to be changed, according to figure 5.36, to include a refer-
ence to CarRegistry. This is needed since Rental will call the method setBookedState
OfCar in CarRegistry.
There is one unhandled issue left, what happens if the car that shall be booked is already
booked? This situation is not handled yet, but should be addressed before the program is
completed. That concludes the implementation of bookCar, the rest of the code can be down-
loaded in the accompanying NetBeans project [Code].

The pay system operation

The final system operation, pay, is designed in figures 5.40 and 5.42. Here, there are two
questions that were left unanswered during design. The first is how the receipt text is created,
the design only shows that a Receipt object is created and passed to the printer. The chosen
solution is to add a method createReceiptString to Receipt, see lines 26-47 in listing
6.36. The printer will call this method to get a a formatted string with the entire receipt text.
This string is then printed to System.out, since there is no real printer in this program. To
create this receipt string, Receipt must gather information from Rental, which leads to the
question what information Rental shall reveal. It has three objects with data, a customer
object, a car object and a payment object. Either we create methods that hand out these
objects, like getRentedCar, or we create methods that hand out the data in the objects, like
getRegNoOfRentedCar. The former alternative, to hand out whole objects, is normally to
prefer. That way objects are kept intact, instead of breaking them up and passing primitive
data. By handing out the whole object, the receiver can call any method in the received object,
not just use its data. This alternative is chosen here, which means, for example, that Receipt
calls rental.getPayment().getTotalCost() to get the cost of the rental.
The other unresolved issue left from design is the method calculateTotalCost in Cash
Payment. Actually, we have no idea how a total cost is calculated. A very simple solution is
chosen, but it would certainly have to be improved if development were to continue to a com-
plete car rental system. This simple solution is to just use the price in CarDTO as total cost,
which means mileage and insurance costs are ignored, and also that there can be no discounts
or campaigns. To implement this solution, CashPayment uses paidRental.getRentedCar().
getPrice() as total cost, line 29 in listing 6.37. This might seem a strange solution. Why
shall Rental call CashPayment, which then only calls back to Rental to get the cost. Why
not just let Rental pass the cost to CashPayment? The reason is that it is assumed that,
as the program grows, CashPayment will have to gather more information, like driven dis-

115
 Chapter 6 Programming

tance and possible discounts. It is CashPayment who has this knowledge about what data is
needed, from where to get it, and how to use it to calculate the total cost.

1 package se.leiflindback.oodbook.rentcar.model;
2
3 import java.time.LocalDateTime;
4
5 /**
6 * The receipt of a rental
7 */
8 public class Receipt {
9 private final Rental rental;
10
11 /**
12 * Creates a new instance.
13 *
14 * @param rental The rental proved by this receipt.
15 */
16 Receipt(Rental rental) {
17 this.rental = rental;
18 }
19
20 /**
21 * Creates a well-formatted string with the entire content
22 * of the receipt.
23 *
24 * @return The well-formatted receipt string.
25 */
26 public String createReceiptString() {
27 StringBuilder builder = new StringBuilder();
28 appendLine(builder, "Car Rental");
29 endSection(builder);
30
31 LocalDateTime rentalTime = LocalDateTime.now();
32 builder.append("Rental time: ");
33 appendLine(builder, rentalTime.toString());
34 endSection(builder);
35
36 builder.append("Rented car: ");
37 appendLine(builder, rental.getRentedCar().getRegNo());
38 builder.append("Cost: ");
39 appendLine(builder, rental.getPayment().
40 getTotalCost().toString());
41 builder.append("Change: ");
42 appendLine(builder, rental.getPayment().
43 getChange().toString());

116
 Chapter 6 Programming

44 endSection(builder);
45
46 return builder.toString();
47 }
48
49 private void appendLine(StringBuilder builder,
50 String line) {
51 builder.append(line);
52 builder.append("\n");
53 }
54
55 private void endSection(StringBuilder builder) {
56 builder.append("\n");
57 }
58 }

Listing 6.36 The class Receipt, after implementing the pay system operation.

1 package se.kth.ict.oodbook.rentcar.model;
2
3 /**
4 * Represents one specific payment for one specific rental. The
5 * rental is payed with cash.
6 */
7 public class CashPayment {
8 private Amount paidAmt;
9 private Amount totalCost;
10
11 /**
12 * Creates a new instance. The customer handed over the
13 * specified amount.
14 *
15 * @param paidAmt The amount of cash that was handed over
16 * by the customer.
17 */
18 public CashPayment(Amount paidAmt) {
19 this.paidAmt = paidAmt;
20 }
21
22 /**
23 * Calculates the total cost of the specified rental.
24 *
25 * @param paidRental The rental for which the customer is
26 * paying.
27 */
28 void calculateTotalCost(Rental paidRental) {

117
 Chapter 6 Programming

29 totalCost = paidRental.getRentedCar().getPrice();
30 }
31
32 /**
33 * @return The total cost of the rental that was paid.
34 */
35 Amount getTotalCost() {
36 return totalCost;
37 }
38
39 /**
40 * @return The amount of change the customer shall have.
41 */
42 Amount getChange() {
43 return paidAmt.minus(totalCost);
44 }
45 }

Listing 6.37 The class CashPayment, after implementing the pay system operation.

6.6 Common Mistakes

Below follows a list of common coding mistakes.

• Incomplete comments. Each public declaration (class, method, etc) shall have a
javadoc comment. Method comments shall cover parameters and return values, us-
ing the javadoc tags @param and @return. It is often argued that it is unnecessary
to comment getter and setter methods. That might very well be the case, but how long
does it take to add a one line comment to a getter or setter? It might even be that the
IDE can generate the comment. If every public declaration has a comment, there is no
risk of missing to comment something by mistake, or by pure laziness.

• Excessive comments. There should be no comments besides the above mentioned NO!
javadoc comments. If there is a need for more comments to explain the code, it proba-
bly means the code is too complex, and has low cohesion.

• Comments written too late. Write the comments together with the code that is com-
mented, maybe even before. That way, writing the comment makes it necessary to
clarify what the code shall do, before (or immediately after) it is written. Also, if com-
ments are written together with the code, they will be of use in future development
of the program. If comments are written last, when the program is already working,
commenting is just a burden, and probably quite a heavy burden.

118
 Chapter 6 Programming

• Many of the common design mistakes can be introduced when programming, even if
they were avoided during design. For example, there is the risk to use primitive data
instead of objects, to use static declarations when they are not appropriate or to place
input or output outside the view. See the text on common design mistakes in section
5.6 for more details on this.
NO!
• Section 6.4, on code smell and refactorings, covers many things that shall be avoided
when coding. Maybe the most common of those possible mistakes are meaningless
names and unnamed values.

119
Chapter 7

Testing
How is it possible to know if a program works?
The answer to that question makes a very big dif-
ference. If it is complicated to verify that the pro-
gram works as intended, developers will be ex-
tremely reluctant to make changes. They will nei-
ther be willing to apply refactorings to improve
the design of existing code, nor to change existing
functionality. Instead they will argue against cus-
tomer’s requirement changes, and solve all prob-
lems by adding new code. This is a disastrous
state of development, characterized by fear, un-
certainty and doubt. Because of the reluctance to
work with existing code, developers will have lit-
Figure 7.1 Lack of tests will bring fear, uncertainty
and doubt, since programmers can not trust the tle knowledge about the code and the code will be
program. Image by unknown creator [Public domain], in bad state. This will make them even more re-
via https://pixabay.com
luctant to make changes, which will lead to even
less knowledge and worsen code state even more. This is exactly the opposite of the flexibility
we want to achieve. The code will constantly become less flexible.
If, on the other hand, it is very easy to ver-
ify that the code works as intended, developers
will be happy to change it. They will constantly
improve its design with refactorings. They will
also be glad to improve customer satisfaction by
adjusting to changing requirements. This is the
flexibility of well-designed software! Code qual-
ity constantly improves, developers gets better
knowledge about the code, and thereby becomes
even more willing to change it.
The difference between the two scenarios
above is automated tests. There should be a test
program which gives input to the program under Figure 7.2 Complete tests will bring confi-
test, and also evaluates the output. If a test passes, dence, since programmers can trust the pro-
gram. Image by unknown creator [Public domain], via
the test program does not do anything. If a test https://pixabay.com

120
 Chapter 7 Testing

fails, it prints an informative message about the failure. With extensive tests that cover all, or
most, possible execution paths through the program with all, or most, possible variable val-
ues, it is guaranteed that the program works if all tests pass. This is a very good situation, one
command starts the test, which tells if the program under test works or, if not, exactly which
problems there are.

7.1 Unit Tests and The JUnit Framework

A unit test is a test of the smallest possible piece of code that makes sence on its own, typically
a method. Unit tests constitute, by far, the most common way for developers to verify that their
code works as intended. Listings 7.1 and 7.2 show a first example of a unit test. The first of
those listings contains the system under test, SUT. It is a method equals, in a class Amount.
The method shall compare two Amount instances and return true if they represent the same
amount, or false if they represent different amounts. Listing 7.2 contain a unit test for that
method. It creates two Amount instances representing the same amount (lines four and five),
calls the equals method (line 7) and verifies that the result is as expected (lines eight to ten).
This test is written using the JUnit 4 framework. It will be covered in more detail below, when
JUnit has been introduced.

1 public boolean equals(Object other) {

2 if (!(other instanceof Amount)) {
3 return false;
4 }
5 Amount otherAmount = (Amount) other;
6 return amount == otherAmount.amount;
7 }

Listing 7.1 The SUT is the equals method in the class

Amount

1 @Test
2 public void testEqual() {
3 int amount = 3;
4 Amount instance = new Amount(amount);
5 Amount other = new Amount(amount);
6 boolean expResult = true;
7 boolean result = instance.equals(other);
8 assertEquals(
9 "Amount instances with same states are not equal.",
10 expResult, result);
11 }

Listing 7.2 A unit test for the code in listing 7.1

121
 Chapter 7 Testing

Frameworks
There are many frameworks that facilitate unit testing, since it is an extremely common testing
approach. JUnit was one of the first, and is also very frequently used. But why use a framework
at all? And exactly what is a framework? A framework provides some functionality that is
not specific for a particular applica-
tion, but is needed in different ap-
plications. Think of the Java APIs
from Oracle, they provide function-
ality for common tasks, and can
be used in many different appli-
cations. In contrast to an API, a
framework not only provides code,
but also flow control. This means
the main method is in the frame-
work, not in application code writ-
ten by application developers. The
framework is responsible for call-
ing application code at the right Figure 7.3 The application fits in the framework like a piece in a
time. This fact, that the application puzzle. Execution, the colored lines, enter the application via
is relieved of flow control responsi- method calls from the framework.
bility, is very important. Consider
for example a framework providing some security control. It would be very hard for applica-
tion developers to always remember, and never forget, to call the framework in all necessary
places. Just one miss would introduce a security hole. If, instead, the framework itself has
the main method and is responsible for when to handle security, application code will be com-
pletely relieved of everything related to security control. This is illustrated in figure 7.3, where
the blue piece, representing the application, is placed inside the framework, represented by all
the white pieces. The colored lines are different executions through the program. Execution
may start in a main method inside the framework, as is the case for the black line. Execution
may also enter the framework from the outside, for example via a network call, as is the case
for the red and green lines. But execution never starts in the application. When the colored
lines enter the application piece, it typically means the framework has called a method in the
application. When the lines exit the application piece, that method has returned.
There are many good reasons to use a framework whenever one can be found. First, a
framework is thoroughly tested and proven to work well. If it did not work well, it would
not be used. Second, if there are many developers using the same framework, there will be
lots of documentation, and it will be easy to get help. Third, the fact that the framework is
responsible for flow control makes sure all code is executed in correct order. Last, not using a
framework means writing new code, which means introducing new bugs.

122
 Chapter 7 Testing

JUnit
JUnit[JU] is one of the most popular unit testing frameworks for Java. It is based on anno-
tations. An annotation is a part of a Java program that is not executed, but instead provides
information about the program for the compiler, or for the JVM, or, as is the case here, for a
framework (JUnit). An annotation is usually used for properties unrelated to the functionality
of the source code, for example to configure security, networking, multithreading or testing.
It starts with the at sign, @, for example @SomeAnnotation. It may take parameters, for
example @SomeAnnotation(someString = "abc", someBoolean = true). When
writing tests with JUnit, annotations are used to specify the content of methods in the test
code, for example that a certain method contains a test. Some of the most common JUnit
annotations are explained in table 7.1

Annotation Example Explanation

@Test aTest contains tests and will be
public void aTest() executed when tests are run.
@Test(expected=Exception.class) Test fails if aTest does not throw an
public void aTest() exception of class Exception
@Ignore(“Not implemented”) aTest will not be executed.
@Test
public void aTest()
@Before prepareTest is executed before each
public void prepareTest() test method.
@After cleanup is executed after each test
public void cleanup() method.
@BeforeClass prepareTests is executed once
public void prepareTests() before the first test in this class.
@AfterClass cleanup is executed once after the last
public void cleanup() test in this class.

Table 7.1 Some of the most common JUnit annotations

A fully automated test must not only call the SUT, but also evaluate if the result of the call
is the expected, that is, if the test passed or failed. This evaluation is done with assert methods
in JUnit. An assert method verifies that its parameters meet some contraint, for example that
they are equal. If the constraint is met, the test passes and nothing is printed to the console. If
the parameters do not meet the constraint, the test fails and the specified explaining message
is printed. Some of the most common assert methods are explained in table 7.2.
With this knowledge about frameworks and JUnit, we can understand the first example in
listing 7.2 in more detail. The complete test class is listed in listing 7.3.

123
 Chapter 7 Testing

Assertion Example Explanation

fail(“explanation”) Always fails. Can be placed at a
code line that should never be
reached.
assertTrue(“explanation”, condition) Passes if condition is true.
assertFalse(“explanation”, condition) Passes if condition is false.
assertEquals(“explanation”, expected, Passes if expected and
actual) actual are equal. expected
and actual can be of any Java
type.
assertNull(“explanation”, object) Passes if object is null.
assertNotNull(“explanation”, object) Passes if object is not null.

Table 7.2 Some of the most common JUnit assert methods

1 package se.kth.ict.oodbook.tests.firstexample;
2
3 import org.junit.After;
4 import org.junit.Before;
5 import org.junit.Test;
6 import static org.junit.Assert.*;
7
8 public class AmountTest {
9 private Amount amtNoArgConstr;
10 private Amount amtWithAmtThree;
11
12 @Before
13 public void setUp() {
14 amtNoArgConstr = new Amount();
15 amtWithAmtThree = new Amount(3);
16 }
17
18 @After
19 public void tearDown() {
20 amtNoArgConstr = null;
21 amtWithAmtThree = null;
22 }
23
24 @Test
25 public void testEqualsNull() {

124
 Chapter 7 Testing

26 Object other = null;

27 boolean expResult = false;
28 boolean result = amtNoArgConstr.equals(other);
29 assertEquals("Amount instance equal to null.",
30 expResult, result);
31 }
32
33 @Test
34 public void testEqualsJavaLangObject() {
35 Object other = new Object();
36 boolean expResult = false;
37 boolean result = amtNoArgConstr.equals(other);
38 assertEquals("Amount instance equal to " +
39 "java.lang.Object instance.",
40 expResult, result);
41 }
42
43 @Test
44 public void testNotEqualNoArgConstr() {
45 int amountOfOther = 2;
46 Amount other = new Amount(amountOfOther);
47 boolean expResult = false;
48 boolean result = amtNoArgConstr.equals(other);
49 assertEquals("Amount instances with different " +
50 "states are equal.",
51 expResult, result);
52 }
53
54 @Test
55 public void testNotEqual() {
56 int amountOfOther = 2;
57 Amount other = new Amount(amountOfOther);
58 boolean expResult = false;
59 boolean result = amtWithAmtThree.equals(other);
60 assertEquals("Amount instances with different " +
61 "states are equal.",
62 expResult, result);
63 }
64
65 @Test
66 public void testEqual() {
67 int amountOfOther = 3;
68 Amount other = new Amount(amountOfOther);
69 boolean expResult = true;
70 boolean result = amtWithAmtThree.equals(other);
71 assertEquals("Amount instances with same states " +

125
 Chapter 7 Testing

72 "are not equal.",

73 expResult, result);
74 }
75
76 }

Listing 7.3 Complete unit test for the equals method in listing 7.1

On line 12 in listing 7.3, the setUp method is annotated @Before. This means it is exe-
cuted before each test method. That way, each test is performed on the two new Amount ob-
jects created in setUp. In a similar way, the tearDown method, which is annotated @After
on line 18, is executed after each test method. That way, the Amount instances on which the
test was performed are dropped, and will not be used for any more test. Each method contain-
ing a test is annotated @Test, see lines 24, 33, 43, 54, and 65. Each of these methods will be
called by JUnit when the tests are executed. All test methods follow the same pattern. First,
they set up the test creating required objects. Second, they define the expected result of the
call to the SUT. Third, the SUT is called and the actual result is saved. Finally, the expected
and actual results are evaluated to check if the test passed. This is a very typical layout of a
test method, but there are other alternatives, as we will see below.

7.2 Unit Testing Best Practices

Much can be said about best practices for
unit tests, but absolutely most important is
to get started writing them. Any test, no mat-
ter what shortcomings it has, is better than no
test at all. Therefore, do not spend so much
time planning tests that the burden of writing
them becomes so big that they are not writ-
ten, or that they cover only a small part of Figure 7.4 Most important is that tests are written. Noth-
the code. Better to start writing and improve ing beats the feeling of seeing all tests pass and
knowing that the program works.
them later, when need is discovered. Below
follows a collection of best practices for unit
testing.
Write tests! Preferably a lot of them. It is essential to have a large and increasing number of
unit tests. To reach this goal, make it a habit never to test anything manually. Whenever
program functionality shall be verified, write a test. Never give input and evaluate output
manually without having first written a unit test. Also never remove a test. If a certain
test seems unnecessary, add an @Ignore annotation instead of deleting it. A test that
at some point seems meaningless might very well later turn out to be useful.
Tests for every known bug When a bug is found, immediately add a test that fails because
of the bug. Only when that test is in place may bug-fixing start. That way, there will
always be a test for everything in the program that we know can go wrong.

126
 Chapter 7 Testing

Do not over-design There is no need to design or document test code as thoroughly as the
product that is tested. Allow a certain amount of hacking when writing tests. In test
code, we can play around a bit and write some of those funny and interesting hacks that
never really seem to fit in production code.

Testing takes time, but it is worth that time A rough estimate, which is true remarkably of-
ten, is that test code has about the same length as the tested code, and takes about the
same time to write. This means it is quite time consuming to write tests. However, it
is also true remarkably often that once the tests are in place, they give immediate return
on the time invested in writing them. This return comes as confidence that the code is
really working, and that it will be easy to verify that it is still working if we have to
make changes.

Independent and self-evaluating To get highest possible value from the tests, they shall be
quick and easy to execute. This means they shall start with one command (or one click in
an IDE), no complex manual setup shall be required. Also, they must be self-evaluating.
Either a test passes, and prints nothing, or it fails and gives a short informative message
about the failure. It must not be required to manually evaluate return values from calls
to the SUT. Finally, all tests must be independent. Do not rely on them being executed
in a specific order, or on previous tests having passed. All test executions must give the
same result.

Organization Place a test class in the same pack-

age as the class it tests. This enables test-
ing of package private methods. However, do
not mix test classes with the SUT. It is better
to maintain two different directory structures,
one for the program itself and one for the tests,
figure 7.5. That way it is easy to see which
code is tests and which is the SUT. It is also
easy to deliver only the product itself, without
tests. These two parallel directory structures
are maintained by all common IDEs, it is not
required to arrange them manually. Figure 7.5 A test class is placed in the same
It is common to write one test class per tested package as the class it tests, but in a differ-
ent directory.
class, and give it the same name as the tested
class, but appending Test to the name. For
example, the tests for a class called Person are in a class called PersonTest. This
makes it easy to find the tests for a certain class. Names of test methods normally
start with test, followed by a description of the test. For example, the method
testNotEqual on line 55 in listing 7.3 tests that the equals method returns false
for two objects that are not equal. It is possible to write any number of assertions in the
same test method, but execution will stop after first failed assertion. Also, test results
are listed per method, individual assertions are not shown. Therefore, it is best to write
few assertions in each method, and instead write more methods.

127
 Chapter 7 Testing

What to test? Test public, protected and package private code, but not private. A private
method can not be tested, since it can not be called from the test class. Also, if methods
with all other accessibilities work, also private methods work. It is normally not needed
to test setters or constructors that only save values, nor getters that only return a value.
Unless bugs appear, we can take for granted that such methods work as intended. Tests
shall cover as much as possible of the code in the SUT. Try to cover all branches of
if statements. Also, try to test boundary conditions and extreme parameter values, like
null, zero, negative values, objects of wrong type etc. It is also important to test that
a method fails the correct way if illegal parameter values are given, or if some other
precondition is not met.
Never worsen SUT design! Try to never, under any condition, worsen the design of the SUT
just to enable testing. This is a slightly controversial statement. It is often suggested,
for example, to break encapsulation by adding get methods to enable retrieving the state
of an object. The only purpose of breaking encapsulation would be to verify that the
state is correct after a method in the object is tested. This is, however, practically never
necessary. Using hard work, a lot of fantasy, and pragmatically testing more than one
method together, we can almost always write tests without worsening SUT design. A
method must have some effect somewhere, otherwise it is useless. To test it, we just
have to find a way to dig out that effect. As an example, consider a class that can create,
read, update and delete values in some storage that can not be accessed by test code.
These methods can be tested together, for example create a value, read it, and check that
the read value equals the created value. Then create a value, delete it and verify that
it can not be read. The create, update, read, etc. More on testing in difficult situations
follows below.

7.3 When Testing is Difficult

Some methods are very complicated to test, but it is
almost always possible! This section covers three dif-
ferent situations when testing might be difficult. One,
it is hard to give input to the SUT. Two, it is hard to
read the test result. Three, the SUT has complex de-
pendencies on other objects and is therefore hard to
start.
Figure 7.6 Testing can be very complicated
Hard to give input The SUT might not get input from and frustrating, but with hard work, a
lot of fantasy, and pragmatically testing
method parameters, but from a file, a database, a more than one method together, we can
complex set of objects or another source. In this almost always write tests without wors-
ening SUT design. Image by unknown creator
case, it is not obvious how the test shall provide [Public domain], via https://pixabay.com
the input. If the SUT reads from file, the test
can write a file with appropriate content. If the SUT reads from a database, the test can
create a database or insert data into an existing database. If the SUT gets data from other
objects, the test can create all objects needed and somehow make them available to the

128
 Chapter 7 Testing

SUT. It is quite common to write a large amount of test code in order to create files,
databases, etc required for testing. Whatever test structure is created, must be deleted
after the test is executed. Remember that tests must be completely independent and
repeatable. A test must leave no traces of its execution.

Hard to read output It might be that no usable result is returned by the tested method, nor
is there any getter that can be used to read the result. In this situation, do never write
a getter to facilitate testing, since it breaks encapsulation of the SUT. Fact is that the
SUT must update something somewhere, or it would be useless. Maybe the problematic
method can not be tested alone, but there is often some combination of method calls that
will show if the test passed. This reasoning is expanded above, in the paragraph labeled
Never worsen SUT design in section 7.2.

Complex dependencies Classes in higher layers depend on

classes in lower layers. The controller in figure 7.7, to the
right, depends on the model, the database integration layer
and the database itself. If a test for the controller fails, we
do not know which of these layers has the bug. A simple
solution to this problem is to write unit tests as usual for all
classes, and let all tests execute code all the way down to
the database. The lowest class with a failed test is the class
with the bug. This is not a pure unit test, since a call to the
controller will execute code also in the model and integra-
tion layers. However, this does not matter very much since
all code is tested and it is possible to locate bugs. What is
more important, is that this strategy leaves the SUT com-
pletely unchanged! Figure 7.7 The SUT might be
hard to test since it depends
Finally, it can not be stated often enough, whatever the problem on many other objects.
is, do not worsen SUT design just to enable testing.

7.4 Unit Testing Case Study

This section does not include a complete listing of all unit tests. That can be found in the ac-
companying NetBeans project, which can be downloaded from GitHub [Code]. Here follows
a description of the first tests that were written, and of tests where particular afterthought was
needed.

NetBeans Support for Unit Testing

NetBeans [NB] is used when developing this unit test case study. This section illustrates how
NetBeans facilitates creating the tests. Similar functionality is available in all major IDEs.

129
 Chapter 7 Testing

To generate a new test class in NetBeans, right-click the project and choose New → Test
for Existing Class.... This will display the New Test For Existing Class di-
alog, which is depicted in figure 7.8. Click the Browse... button, marked with a red cir-
cle, to choose for which class tests shall be generated. In this example, the chosen class is
CarRegistry, from the rent car case study. Last, click Finish, and NetBeans will generate
test code similar to listing 7.4.

Figure 7.8 NetBeans’ New Test For Existing Class dialog. The Browse... button, marked with a
red circle, is used to decide for which class tests shall be generated.

1 package se.kth.ict.oodbook.rentcar.integration;
2
3 import org.junit.After;
4 import org.junit.AfterClass;
5 import org.junit.Before;
6 import org.junit.BeforeClass;
7 import org.junit.Test;
8 import static org.junit.Assert.*;
9
10 public class CarRegistryTest {
11 @BeforeClass
12 public static void setUpClass() {

130
 Chapter 7 Testing

13 }
14
15 @AfterClass
16 public static void tearDownClass() {
17 }
18
19 @Before
20 public void setUp() {
21 }
22
23 @After
24 public void tearDown() {
25 }
26
27 @Test
28 public void testFindAvailableCar() {
29 System.out.println("findAvailableCar");
30 CarDTO searchedCar = null;
31 CarRegistry instance = new CarRegistry();
32 CarDTO expResult = null;
33 CarDTO result = instance.findAvailableCar(searchedCar);
34 assertEquals(expResult, result);
35 // TODO review the generated test code and remove the
36 // default call to fail.
37 fail("The test case is a prototype.");
38 }
39
40 @Test
41 public void testSetBookedStateOfCar() {
42 System.out.println("setBookedStateOfCar");
43 CarDTO car = null;
44 CarRegistry instance = new CarRegistry();
45 instance.setBookedStateOfCar(car, false);
46 // TODO review the generated test code and remove the
47 // default call to fail.
48 fail("The test case is a prototype.");
49 }
50 }

Listing 7.4 Skeleton code for a test class, generated by NetBeans.

The class has the same name as the tested class, but with Test appended to the
class name (line 10). All four before and after methods are generated (lines 11-
25), but they are empty. If some code is needed to prepare a test or to clean up
after a test, it shall be added here. Methods that remain empty can be removed.
One test method is generated for each public, protected or package private method in

131
 Chapter 7 Testing

the SUT (lines 27-49). These methods contain a printout (lines 29 and 42), which
should be removed since tests are not supposed to produce any output if they pass.
After this, the test methods create an instance
of the SUT (lines 31 and 44) and of other
objects that are necessary to perform the test
(lines 30 and 43). There is no guarantee these
objects are created correctly, always check if
changes are required. Next, the tested method
is called and the result is saved in a vari-
able (lines 33 and 45). Then an assertion is
called to evaluate the test result (line 34). The
testSetBookedStateOfCar method contains
no assertion, since setBookedStateOfCar is
void. In this case, NetBeans does not know how
to evaluate the outcome. Again. even if the asser-
tion is generated, there is no guarantee it is cor-
rect. Finally, there is a TODO comment and a call
to fail (lines 35-37 and 46-48), which should
both be removed when the test is completed.
To execute the tests, right-click the NetBeans
project and chose Test, as depicted in figure 7.9.
The test result will be displayed in a window sim-
ilar to figure 7.10. Currently, none of the two
Figure 7.9 To run the tests, right-click the Net-
auto-generated tests, testFindAvailableCar
Beans project and chose Test. and testSetBookedStateOfCar, are imple-
mented. As a result, they both fail.

Figure 7.10 NetBeans’ test result window. The auto-generated code always makes a test fail.

132
 Chapter 7 Testing

Writing the Tests

Tests will be written in bottom-up order, first for classes with no dependencies on other classes,
then for classes with dependencies. This means we will only write tests for classes without
dependencies, or for classes that are already tested. That way, it becomes possible to run a test
as soon as it is written, and immediately know if the tested class works as intended.
It must be emphasized that the workflow followed here is very unnatural. We first wrote
the entire program and then started to write tests. Normally, a test is written either before or
immediately after the method that shall be tested. The only reason for this workflow is to mix
!
theory and practice, by using programming best practices as soon as they were introduced.
This odd workflow means the first task is to identify a class with no dependency on any other
class. A natural place to start looking for such a class is in the lowest layer, integration.
That layer contains CarDTO, which depends on Amount; CarRegistry, which depends
on CarDTO; RegistryCreator, which depends on CarRegistry and RentalRegistry;
RentalRegistry, which depends on Rental; and Printer, which depends on Receipt.
No suitable class was found in the integration layer, next candidates will be classes on
which the classes in the integration layer depends. The first that was mentioned was
Amount, which in fact has no dependency. It will be the first class to test.

The First Tested Class, Amount

All public, protected and package private methods shall be tested, but not private. The Amount
class has only public methods and constructors, so they shall all be tested. The constructors,
however, contain no logic, they only set a value. Therefore, they will not be tested. The first
method that is tested is equals, since other tests will use it, as will soon be clear. The equals
method is listed in listing 7.5.

1 /**
2 * Two <code>Amount</code>s are equal if they represent the
3 * same amount.
4 *
5 * @param other The <code>Amount</code> to compare with this
6 * amount.
7 * @return <code>true</code> if the specified amount is equal
8 * to this amount, <code>false</code> if it is not.
9 */
10 @Override
11 public boolean equals(Object other) {
12 if (other == null | | !(other instanceof Amount)) {
13 return false;
14 }
15 Amount otherAmount = (Amount) other;
16 return amount == otherAmount.amount;
17 }

Listing 7.5 The equals method of the Amount class

133
 Chapter 7 Testing

To take the branch other == null, on line 12, the method must be called with a
null parameter. To take the !(other instanceof Amount) branch, also on line 12,
the parameter must be an object that is not an instance of Amount. An easy choice
is to use a java.lang.Object instance. Finally, there are two different executions
of line 16, one where amount == otherAmount.amount and one where amount !=
otherAmount.amount. These tests can be found in listing 7.6.

1 package se.kth.ict.oodbook.rentcar.model;
2
3 import org.junit.After;
4 import org.junit.Before;
5 import org.junit.Test;
6 import static org.junit.Assert.*;
7
8 public class AmountTest {
9 private Amount amtNoArgConstr;
10 private Amount amtWithAmtThree;
11
12 @Before
13 public void setUp() {
14 amtNoArgConstr = new Amount();
15 amtWithAmtThree = new Amount(3);
16 }
17
18 @After
19 public void tearDown() {
20 amtNoArgConstr = null;
21 amtWithAmtThree = null;
22 }
23
24 @Test
25 public void testEqualsNull() {
26 Object other = null;
27 boolean expResult = false;
28 boolean result = amtNoArgConstr.equals(other);
29 assertEquals("Amount instance equal to null.",
30 expResult, result);
31 }
32
33 @Test
34 public void testEqualsJavaLangObject() {
35 Object other = new Object();
36 boolean expResult = false;
37 boolean result = amtNoArgConstr.equals(other);
38 assertEquals("Amount instance equal to " +
39 "java.lang.Object instance.",

134
 Chapter 7 Testing

40 expResult, result);
41 }
42
43 @Test
44 public void testNotEqualNoArgConstr() {
45 int amountOfOther = 2;
46 Amount other = new Amount(amountOfOther);
47 boolean expResult = false;
48 boolean result = amtNoArgConstr.equals(other);
49 assertEquals("Amount instances with different states" +
50 " are equal.", expResult, result);
51 }
52
53 @Test
54 public void testNotEqual() {
55 int amountOfOther = 2;
56 Amount other = new Amount(amountOfOther);
57 boolean expResult = false;
58 boolean result = amtWithAmtThree.equals(other);
59 assertEquals("Amount instances with different states" +
60 " are equal.", expResult, result);
61 }
62
63 @Test
64 public void testEqual() {
65 int amountOfOther = 3;
66 Amount other = new Amount(amountOfOther);
67 boolean expResult = true;
68 boolean result = amtWithAmtThree.equals(other);
69 assertEquals("Amount instances with same states are" +
70 " not equal.", expResult, result);
71 }
72 }

Listing 7.6 Tests for all possible paths through the equals method of the Amount class

Next thing to look for is extreme values of the parameters. All obvious extreme values,
like null, are already covered. However, since this is our first test, we might be extra careful
and test also with an Amount object representing zero. That way, also the default constructor
will be executed in a test. Theoretically, we could test with Amounts representing positive,
negative, Integer.MAX_VALUE and Integer.MIN_VALUE amounts, but there is really no
reason to suspect that the method would behave differently for such values. Listing 7.7 shows
the test for an Amount with the value zero.

135
 Chapter 7 Testing

1 @Test
2 public void testEqualNoArgConstr() {
3 int amountOfOther = 0;
4 Amount other = new Amount(amountOfOther);
5 boolean expResult = true;
6 boolean result = amtNoArgConstr.equals(other);
7 assertEquals("Amount instances with same states are" +
8 " not equal.", expResult, result);
9 }

Listing 7.7 Test for the equals method of an Amount representing the amount zero.

Finally, is there any way the parameter can have an illegal value, or is there some precon-
dition that must be met for method to work properly? The answer is “no, the method should
function the same way for all possible parameter values”. That means we are done testing it.
Remember to run the tests and check that they all pass, figure 7.11. Our first green bar!!

Figure 7.11 All tests for the equals method pass.

The other Amount methods, namely minus, plus and toString, are independent, neither
any of the methods, nor its test, will use any of the other methods. Therefore, they can be
written in any order. Let’s start with minus, which is listed in listing 7.8.

1 /**
2 * Subtracts the specified <code>Amount</code> from this
3 * object and returns an <code>Amount</code> instance with
4 * the result.
5 *
6 * @param other The <code>Amount</code> to subtract.
7 * @return The result of the subtraction.
8 */
9 public Amount minus(Amount other) {
10 return new Amount(amount - other.amount);
11 }

Listing 7.8 The minus method of the Amount class

136
 Chapter 7 Testing

This method has only one execution path, since there are no flow control statements. There
are no illegal parameter values, but the subtraction may overflow. If, for example, -1 is sub-
tracted from Integer.Min_VALUE, the result is a negative integer with a magnitude too big
to fit in an int. In fact, we have discovered a flaw in the design. The method ought to check
if an overflow occurred, and, if so, throw an exception. However, since exception handling is
covered later in a later chapter, this check is not introduced here. Instead, an explaining text
is added to the javadoc comment, saying that The operation will overflow if the
result is smaller than <code>Integer.MIN_VALUE</code>. What can then be
tested regarding overflow? Nothing in fact, the method might fail, but the failure is not han-
dled in any way. The conclusion is that one test would probably be enough, just perform a
subtraction and check that the result is correct. However, since we have just started, let’s be a
bit overambitious and test positive, negative and zero results, see listing 7.9. Once the first test
for minus is written, it takes about thirty seconds to add the other two, and the more tests that
pass, the greater the pleasure to see them pass. Note that assertEquals, which is called on
lines 10, 23 and 35, will use the equals method in Amount to verify that the two specified
Amount instances are equal. This is why it was important to know that equals worked when
minus was tested. It is now clear that if a test for minus fails, it is because of a bug in minus,
not in equals.

1 @Test
2 public void testMinus() {
3 int amountOfOperand1 = 10;
4 int amountOfOperand2 = 3;
5 Amount operand1 = new Amount(amountOfOperand1);
6 Amount operand2 = new Amount(amountOfOperand2);
7 Amount expResult = new Amount(amountOfOperand1 -
8 amountOfOperand2);
9 Amount result = operand1.minus(operand2);
10 assertEquals("Wrong subtraction result",
11 expResult, result);
12 }
13
14 @Test
15 public void testMinusNegResult() {
16 int amountOfOperand1 = 3;
17 int amountOfOperand2 = 10;
18 Amount operand1 = new Amount(amountOfOperand1);
19 Amount operand2 = new Amount(amountOfOperand2);
20 Amount expResult = new Amount(amountOfOperand1 -
21 amountOfOperand2);
22 Amount result = operand1.minus(operand2);
23 assertEquals("Wrong subtraction result",
24 expResult, result);
25 }
26 @Test

137
 Chapter 7 Testing

27 public void testMinusZeroResultNegOperand() {

28 int amountOfOperand1 = -3;
29 int amountOfOperand2 = -3;
30 Amount operand1 = new Amount(amountOfOperand1);
31 Amount operand2 = new Amount(amountOfOperand2);
32 Amount expResult = new Amount(amountOfOperand1 -
33 amountOfOperand2);
34 Amount result = operand1.minus(operand2);
35 assertEquals("Wrong subtraction result",
36 expResult, result);
37 }

Listing 7.9 The tests for the minus method of the Amount class

The tests for plus are created exactly the same way as the tests for minus, and are therefore
not covered here. Finally, there is the toString method, which returns a string representa-
tion of the amount, listing 7.10. Also this method is tested with positive, negative, and zero
amounts, see listing 7.11. That concludes testing Amount. There are 15 tests in total, and all
pass, brilliant!

1 @Override
2 public String toString() {
3 return Integer.toString(amount);
4 }
Listing 7.10 The toString method of the
Amount class

1 @Test
2 public void toStringPosAmt() {
3 int representedAmt = 10;
4 Amount amount = new Amount(representedAmt);
5 String expResult = Integer.toString(representedAmt);
6 String result = amount.toString();
7 assertEquals("Wrong string returned by toString",
8 expResult, result);
9 }
10
11 @Test
12 public void toStringNegAmt() {
13 int representedAmt = -10;
14 Amount amount = new Amount(representedAmt);
15 String expResult = Integer.toString(representedAmt);
16 String result = amount.toString();
17 assertEquals("Wrong string returned by toString",
18 expResult, result);
19 }

138
 Chapter 7 Testing

20
21 @Test
22 public void toStringZeroAmt() {
23 int representedAmt = 0;
24 Amount amount = new Amount(representedAmt);
25 String expResult = Integer.toString(representedAmt);
26 String result = amount.toString();
27 assertEquals("Wrong string returned by toString",
28 expResult, result);
29 }

Listing 7.11 The tests for the toString method of the Amount class

The First Problematic Test, a void Method

Following the same reasoning as when testing Amount, tests are written also for CarDTO.
This is quite straightforward, just remember that object parameters must be tested with null,
and string parameters also with an empty string. The code for these tests can be found in
the accompanying NetBeans project [Code]. The next class to test is CarRegistry, which
is a bit more challenging since it has a void method, namely setBookedStateOfCar.
Remember the strategy, the method must have some effect somewhere, just locate that ef-
fect. Here the effect is that a booked car will not be returned by findAvailableCar,
even if the description matches. Therefore, setBookedStateOfCar can be tested to-
gether with findAvailableCar, as illustrated in listing 7.13. Listing 7.12 shows
setBookedStateOfCar and findAvailableCar. There is an interesting problem here,
the calls to assertEquals, for example on line nine in listing 7.13, will use the equals
method in CarDTO to evaluate if two instances are equal. But there is no such method, which
means the default instance of equals, in java.lang.Object, will be used! That method
considers two objects to be equal only if they are exactly the same object, residing in the same
memory location. Since this is not appropriate here, an equals method is added to CarDTO,
and of course it is also tested. It could be argued that the SUT is now changed, only to facil-
itate testing. That might be the case, but what really matters is that the design of the SUT is
not worsened. Furthermore, an equals method might very well turn out to be appropriate for
the SUT itself.

1 /**
2 * Search for a car matching the specified search criteria.
3 *
4 * @param searchedCar This object contains the search criteria.
5 * Fields in the object that are set to
6 * <code>null</code> or <code>0</code> are
7 * ignored.
8 * @return <code>true</code> if a car with the same features
9 * as <code>searchedCar</code> was found,

139
 Chapter 7 Testing

10 * <code>false</code> if no such car was found.

11 */
12 public CarDTO findAvailableCar(CarDTO searchedCar) {
13 for (CarData car : cars) {
14 if (matches(car, searchedCar) && !car.booked) {
15 return new CarDTO(car.regNo, new Amount(car.price),
16 car.size, car.AC, car.fourWD,
17 car.color);
18 }
19 }
20 return null;
21 }
22
23 /**
24 * If there is an existing car with the registration number of
25 * the specified car, set its booked property to the specified
26 * value. Nothing is changed if the car’s booked property
27 * already had the specified value.
28 *
29 * @param car The car that shall be marked as booked.
30 * @param bookedState The new value of the booked property.
31 */
32 public void setBookedStateOfCar(CarDTO car,
33 boolean bookedState) {
34 CarData carToBook = findCarByRegNo(car);
35 carToBook.booked = bookedState;
36 }

Listing 7.12 The setBookedStateOfCar and findAvailableCar methods of the

CarRegistry class

1 @Test
2 public void testSetBookedStateOfCar() {
3 CarDTO bookedCar = new CarDTO("abc123", new Amount(1000),
4 "medium", true, true, "red");
5 CarRegistry instance = new CarRegistry();
6 instance.setBookedStateOfCar(bookedCar, true);
7 CarDTO expResult = null;
8 CarDTO result = instance.findAvailableCar(bookedCar);
9 assertEquals("Booked car was found", expResult, result);
10 }

Listing 7.13 The test for the setBookedStateOfCar method of the CarRegistry class

140
 Chapter 7 Testing

More Difficult Tests

No tests are needed for CustomerDTO, AddressDTO or DrivingLicenseDTO, since they
contain only setters, getters and constructors that do nothing but save or return values.
CashRegister and RentalRegistry can, in fact, not be properly tested. They contain
one void method each, that only updates a field in the same object. This is not a proof that
there are untestable methods, instead, it shows that the program is not complete. When devel-
oping continues, it will certainly be possible to read the balance of the cash register and to see
which rentals have been made. Thereby, it will also possible to test those classes. For now,
however, all that can be done is to call those methods in their tests, there is no way to evaluate
their outcome. The next class that is tested is RegistryCreator, tests are available in the
accompanying NetBeans project [Code].
After that it is not possible to continue in pure bottom-up order any more, since there are no
remaining classes without dependencies or depending only on tested classes. The only thing
to do is to write tests for all remaining classes in model and integration, and then run all
of them. Writing tests for these classes clearly shows, as was already known, that there is no
error handling in this program. For example, there are many methods which should check that
they are not called with null parameters, but they do not. This lack of error handling makes
it meaningless to test those erroneous conditions. That must be postponed until later, when
error handling is added.
The test environment for some methods, for example createReceiptString in Receipt
(listing 7.14), require quite a lot of work to set up. This is quite normal, and should be expected
to happen. Listing 7.15 shows testCreateReceiptString, which illustrates that with
some fantasy and quite a lot of code, it is possible to test also methods depending on many
other objects or methods. Note that the string that makes up the expected result (line 19-24)
does not, in any way, depend on createReceiptString or any other method in Receipt.
This eliminates the risk that the test has the same bug as the SUT.

1 /**
2 * Creates a well-formatted string with the entire content of
3 * the receipt.
4 *
5 * @return The well-formatted receipt string.
6 */
7 public String createReceiptString() {
8 StringBuilder builder = new StringBuilder();
9 appendLine(builder, "Car Rental");
10 endSection(builder);
11
12 LocalDateTime rentalTime = LocalDateTime.now();
13 builder.append("Rental time: ");
14 appendLine(builder, rentalTime.toString());
15 endSection(builder);
16
17 builder.append("Rented car: ");

141
 Chapter 7 Testing

18 appendLine(builder, rental.getRentedCar().getRegNo());
19 builder.append("Cost: ");
20 appendLine(builder, rental.getPayment().getTotalCost().
21 toString());
22 builder.append("Change: ");
23 appendLine(builder, rental.getPayment().getChange().
24 toString());
25 endSection(builder);
26
27 return builder.toString();
28 }
29
30 private void appendLine(StringBuilder builder, String line) {
31 builder.append(line);
32 builder.append("\n");
33 }
34
35 private void endSection(StringBuilder builder) {
36 builder.append("\n");
37 }

Listing 7.14 The createReceiptString method of the Receipt class, and its private
helper methods.

1 @Test
2 public void testCreateReceiptString() {
3 Amount price = new Amount(100);
4 String regNo = "abc123";
5 String size = "medium";
6 boolean AC = true;
7 boolean fourWD = true;
8 String color = "red";
9 CarDTO rentedCar = new CarDTO(regNo, price, size, AC,
10 fourWD, color);
11 Amount paidAmt = new Amount(500);
12 CashPayment payment = new CashPayment(paidAmt);
13 Rental paidRental = new Rental(null, new RegistryCreator().
14 getCarRegistry());
15 paidRental.setRentedCar(rentedCar);
16 paidRental.pay(payment);
17 Receipt instance = new Receipt(paidRental);
18 LocalDateTime rentalTime = LocalDateTime.now();
19 String expResult = "\n\nRented car: " + regNo +
20 "\nCost: " + price +
21 "\nChange: " + paidAmt.minus(price) +
22 "\n\n";

142
 Chapter 7 Testing

23 String result = instance.createReceiptString();

24 assertTrue("Wrong printout.", result.contains(expResult));
25 assertTrue("Wrong rental year.", result.contains(
26 Integer.toString(rentalTime.getYear())));
27 assertTrue("Wrong rental month.", result.contains(
28 Integer.toString(rentalTime.getMonthValue())));
29 assertTrue("Wrong rental day.", result.contains(
30 Integer.toString(rentalTime.getDayOfMonth())));
31 assertTrue("Wrong rental hour.", result.contains(
32 Integer.toString(rentalTime.getHour())));
33 assertTrue("Wrong rental minute.", result.contains(
34 Integer.toString(rentalTime.getMinute())));
35 assertTrue("Wrong receipt content.",
36 result.contains(expResult));
37 }

Listing 7.15 The test for the createReceiptString method of the Receipt class

Testing User Interface

printReceipt in Printer (listing 7.16) is an interesting method. It is void, but calling it
has an effect, though only on the screen. It produces output to System.out. Luckily, it is
easy to test such output, since it is possible to replace the stream System.out with another
stream, that prints to a buffer in memory instead of the screen. This is done on line seven
in listing 7.17. The content of this in-memory buffer becomes the outcome of the SUT call,
which is compared with the expected result on line 43.
This is the only user interface test that is written, since development of user interfaces is not
included in the course. However, testing System.in should be done exactly the same way
as testing System.out, by reassigning the stream and let it read from an in-memory buffer
instead of the keyboard. This strategy only works for command line user interfaces, not for
graphical or web-based user interface. Still, it is very much possible to test also such user
interfaces, since there are many frameworks which makes it possible to give input to, and read
output from, different kinds of UIs.

1 /**
2 * Prints the specified receipt. This dummy implementation
3 * prints to <code>System.out</code> instead of a printer.
4 *
5 * @param receipt
6 */
7 public void printReceipt(Receipt receipt) {
8 System.out.println(receipt.createReceiptString());
9 }

Listing 7.16 The outcome of the printReceipt method appears only in System.out.

143
 Chapter 7 Testing

1 public class PrinterTest {

2 private ByteArrayOutputStream outContent;
3 private PrintStream originalSysOut;
4
5 @Before
6 public void setUpStreams() {
7 originalSysOut = System.out;
8 outContent = new ByteArrayOutputStream();
9 System.setOut(new PrintStream(outContent));
10 }
11
12 @After
13 public void cleanUpStreams() {
14 outContent = null;
15 System.setOut(originalSysOut);
16 }
17
18 @Test
19 public void testCreateReceiptString() {
20 Amount price = new Amount(1000);
21 String regNo = "abc123";
22 String size = "medium";
23 boolean AC = true;
24 boolean fourWD = true;
25 String color = "red";
26 CarDTO rentedCar = new CarDTO(regNo, price, size, AC,
27 fourWD, color);
28 Amount paidAmt = new Amount(5000);
29 CashPayment payment = new CashPayment(paidAmt);
30 Rental paidRental = new Rental(null,
31 new RegistryCreator().
32 getCarRegistry());
33 paidRental.setRentedCar(rentedCar);
34 paidRental.pay(payment);
35 Receipt receipt = new Receipt(paidRental);
36 Printer instance = new Printer();
37 instance.printReceipt(receipt);
38 LocalDateTime rentalTime = LocalDateTime.now();
39 String expResult = "\n\nRented car: " + regNo
40 + "\nCost: " + price
41 + "\nChange: "
42 + paidAmt.minus(price) + "\n\n\n";
43 String result = outContent.toString();
44 assertTrue("Wrong printout.",
45 result.contains(expResult));
46 assertTrue("Wrong rental year.", result.contains(

144
 Chapter 7 Testing

47 Integer.toString(rentalTime.getYear())));
48 assertTrue("Wrong rental month.", result.contains(
49 Integer.toString(rentalTime.
50 getMonthValue())));
51 assertTrue("Wrong rental day.", result.contains(
52 Integer.toString(rentalTime.
53 getDayOfMonth())));
54 assertTrue("Wrong rental hour.", result.contains(
55 Integer.toString(rentalTime.getHour())));
56 assertTrue("Wrong rental minute.", result.contains(
57 Integer.toString(rentalTime.getMinute())));
58 }
59 }

Listing 7.17 The test for the createReceiptString method of the Receipt class

The last classes, Controller and Main

As was mentioned above, user interface testing is not included in the course. Thus, the only
remaining classes are Controller and Main. The tests for Controller once again re-
veals the lack of a possibility to read from the rental registry, it is impossible to check any
property of the rental that is created by the Controller methods. In fact, a “read-only reg-
istry” is a very strange thing, why store anything in the registry if it can not later be read? It
is impossible to claim that the design is worsened by a read method in RentalRegistry.
Rather, it is a bug that there is no such method. According to this reasoning, the method
findRentalByCustomerName is added to the RentalRegistry. It returns all rentals made
by a customer with the specified name. Having added this method, the test of saveRental
in RentalRegistry can be extended to verify that the Rental is actually saved. According
to the same reasoning, the method getRentingCustomer is added to Rental. What point
is there to store information about the renting customer, if that information can not be read?
Testing the controller is a bit tricky. Since it is high up in the layer stack, both setting
up test environment, giving input, and reading output, involves many other objects. As an
example, consider the method testRentalWithBookedCarIsStored on line 26 in listing
7.19, which tests that the method bookCar (listing 7.18) correctly stores the current rental to
the rental registry. First, lines 27-36 creates objects that are required input to the SUT. Next,
line 37 prepares the SUT. Only after this call can the tested method, bookCar be called. Lines
39-41 and 47 extracts the result, namely the stored Rental object. The assertEqual call
on lines 44-46 assures that the correct number of rentals (one) is stored. If this is not the case,
there is no pint in continuing the test, it has already failed. Lines 48-54 checks that the correct
rental was stored in the registry. The only way to do this is to print the receipt and check
that the rented car is specified there. A more straightforward way would have been to get the
rented car by calling getRentedCar in Rental, but that method can not be reached by the
test since it is package private. We do not want to worsen the design by making it public, and
thus part of the public interface.

145
 Chapter 7 Testing

1 /**
2 * Books the specified car. After calling this method, the car
3 * can not be booked by any other customer. This method also
4 * permanently saves information about the current rental.
5 *
6 * @param car The car that will be booked.
7 */
8 public void bookCar(CarDTO car) {
9 rental.setRentedCar(car);
10 rentalRegistry.saveRental(rental);
11 }

Listing 7.18 The bookcar method of the Controller class.

1 public class ControllerTest {

2 private Controller instance;
3 private RegistryCreator regCreator;
4 ByteArrayOutputStream outContent;
5 PrintStream originalSysOut;
6
7 @Before
8 public void setUp() {
9 originalSysOut = System.out;
10 outContent = new ByteArrayOutputStream();
11 System.setOut(new PrintStream(outContent));
12 Printer printer = new Printer();
13 regCreator = new RegistryCreator();
14 instance = new Controller(regCreator, printer);
15 }
16
17 @After
18 public void tearDown() {
19 outContent = null;
20 System.setOut(originalSysOut);
21 instance = null;
22 regCreator = null;
23 }
24
25 @Test
26 public void testRentalWithBookedCarIsStored() {
27 String customerName = "custName";
28 CustomerDTO rentingCustomer =
29 new CustomerDTO(customerName,
30 new AddressDTO("street", "zip",
31 "city"),

146
 Chapter 7 Testing

32 new DrivingLicenseDTO("1234567"));
33 String regNo = "abc123";
34 CarDTO rentedCar = new CarDTO(regNo, new Amount(1000),
35 "medium", true,
36 true, "red");
37 instance.registerCustomer(rentingCustomer);
38 instance.bookCar(rentedCar);
39 List<Rental> savedRentals =
40 regCreator.getRentalRegistry().
41 findRentalByCustomerName(customerName);
42 int expectedNoOfStoredRentals = 1;
43 int noOfStoredRentals = savedRentals.size();
44 assertEquals("Wrong number of stored rentals.",
45 expectedNoOfStoredRentals,
46 noOfStoredRentals);
47 Rental savedRental = savedRentals.get(0);
48 Amount paidAmt = new Amount(5000);
49 CashPayment payment = new CashPayment(paidAmt);
50 savedRental.pay(payment);
51 savedRental.printReceipt(new Printer());
52 String result = outContent.toString();
53 assertTrue("Saved rental does not contain rented car",
54 result.contains(regNo));
55 }
56 }

Listing 7.19 The test for the bookcar method of the Controller class

The last class is Main, which has only the method main. This is very hard to test, since it
does nothing but create some objects. When the program is ready, it will most likely start a
user interface, then it will be possible to verify that something happens on the screen. While
this can not be done now, since no user interface is created, it is still possible to verify that
some chosen part of the output from the test run in View.sampleExecution appears on the
screen. It is also possible to inspect the JVM to see that the expected objects are created, but
this involves starting a debugger in another JVM, and attaching it to the inspected JVM, which
is too complicated for this course. Maybe even too complicated to be meaningful at all, if the
only purpose is to see that some new statements behave as expected.
That concludes the rent car testing case study. A total of 56 test methods were created,
which should be acceptable for such a small program, completely lacking exception handling.
All 56 tests pass, which gives the joyful sight presented in figure 7.12. Quite amazingly, the
SUT consists of 1353 lines of code in total, and the tests of 1322 (no cheating). A difference
of only two percent!

147
 Chapter 7 Testing

Figure 7.12 All 56 tests of the case study pass.

7.5 Common Mistakes

Below follows a list of common mistakes made when writing unit tests.

Too few tests Both the most common and most severe mistake is probably not to write
enough tests. Try to cover all possible branches of if statements and loops. Also
NO!
try to write tests for extreme and illegal parameter values.

Too many assertions in the same test method Place as few assertions as possible in each
test method. It is clearer what happens and easier to give an explaining name to a test
method if it has few assertions. Ideally, there should only be one assertion per test
method, but it is not always possible to evaluate the outcome of a call to the SUT in
one single assertion. Sometimes more than one are actually required.

Not self-evaluating Test result should be evaluated using assertions, not with if statements
in the test methods, nor by forcing the tester to read output.
NO!
Producing output A test shall not write to System.out. The more tests there are, the
more confusing it becomes if they print some kind of status messages.

Worsen SUT design The design of the SUT shall not be worsened just to facilitate testing.
It is practically always possible to test without changing the SUT, even though it often
requires extra work.

148
Chapter 8

Handling Failure
A program is not complete if it does not
handle all possible failures. Some failures
have cures, and can be dealt with carefully
in order to make the program work despite
the exceptional condition that caused the
failure. In other cases, there is not much
more to do, than to report that the opera-
tion failed. In any case, the outcome of a
system operation shall never be undefined,
no matter what happens.
Many programming languages enables
using exceptions for error handling, which
is an important mechanism to make the
code more flexible and easier to under-
stand. An exception represents an ab-
normal situation, that disrupts the sequen-
tial execution path through the program.
Say, for example, that a method requires Figure 8.1 A program must handle abnormal situations,
giving the user appropriate feedback. Image by FreeIm-
a connection to a server to fulfill its task, ages.com/Alexandre Galant
and that this connection can not be estab-
lished. This means the method can not do what it is supposed to, and has to return immedi-
ately, informing the caller that the abnormal situation could not connect to server occurred.
The caller then has to switch to error handling, since it did not get any result from the called
method. This scenario is quite easily implemented using exceptions. Without them, it would
instead require messy if statements and return values, in order to check for possible error
codes.

8.1 UML
It is remarkably unclear and difficult to illustrate exception handling in UML. Figure 8.2 shows
two possible ways to draw exception handling in a class diagram. The curly brackets in figure
8.2a define a constraint, which means some condition or restriction related to the element
where it is placed. The content of a constraint is free text, anything can be written there.

149
 Chapter 8 Handling Failure

This particular constraint specifies exceptions that the method may throw. Figure 8.2b, shows
another way to illustrate exceptions in a class diagram, using a reference to the exception class.
It has the disadvantage of not showing which method throws the exception.

(a)

(b)

Figure 8.2 Exception handling in class diagram.

(a) using constraint
(b) using reference to exception class

Figure 8.3 illustrates how a sequence and a communication diagram can indicate that an
exception is thrown, interrupting the normal, sequential, flow. An open arrow is used, see
figure 8.3a, which means the message is asynchronous, and does not follow the sequential
flow. The stereotype «exception» is used to further clarify that the asynchronous message is
an exception. A stereotype says that an element belongs to a certain category of such elements.
Here, it says that the message belongs to the category thrown exception. Unfortunately, astah
can not draw an asynchronous message in a communication diagram. Therefore, a normal,
closed, arrow symbolizing a sequential message is used in the red circle in figure 8.3b. Note
that, at the blue arrow in the sequence diagram, there is an extra horizontal line in the activation
bar. This is because there is a new block starting here, to handle the message symbolizing the
exception. This block is not shifted to the right as usual, since the message is asynchronous.
It can be considered to represent the catch block. Finally, note the unfortunate fact that there
is a parenthesis, (), after the exception name in both diagrams.

(b)
(a)

Figure 8.3 exception handling

(a) in sequence diagram, (b) in communication diagram

150
 Chapter 8 Handling Failure

8.2 Exception Handling Best Practices

This section introduces best practices for exception handling. They should all be considered,
and if some of them are not followed, it should be clear why that is the case. The best practices
are introduced by implementing failure handling for a particular condition. This condition is
when, in the rent car case study, the user tries to book a car, and the booking procedure
fails. That can happen either because the car was already booked, or because the call to the
underlying storage (for example text file or database) failed.
The bookCar system operation is handled by the bookCar method in Controller, see
listing 8.1. To book the car, this method calls setRentedCar in Rental, listing 8.2, which, in
turn, calls setBookedStateOfCar in CarRegistry, listing 8.3. This last method changes
the car’s booked state. As can be seen in these listings, a car is booked without checking
whether it is already booked. This is because no error handling has yet been implemented, but
such a check will have to be added now.

1 /**
2 * Books the specified car. After calling this method, the car
3 * can not be booked by any other customer. This method also
4 * permanently saves information about the current rental.
5 *
6 * @param car The car that will be booked.
7 */
8 public void bookCar(CarDTO car) {
9 rental.setRentedCar(car);
10 rentalRegistry.saveRental(rental);
11 }

Listing 8.1 The bookCar method in Controller

1 /**
2 * Specifies the car that was rented.
3 *
4 * @param rentedCar The car that was rented.
5 */
6 public void setRentedCar(CarDTO rentedCar) {
7 this.rentedCar = rentedCar;
8 carRegistry.setBookedStateOfCar(rentedCar, true);
9 }

Listing 8.2 The setRentedCar method in Rental

1 /**
2 * If there is an existing car with the registration number
3 * of the specified car, set its booked property to the

151
 Chapter 8 Handling Failure

4 * specified value. Nothing is changed if the car’s booked

5 * property already had the specified value.
6 *
7 * @param car The car that shall be marked as booked.
8 * @param bookedState The new value of the booked property.
9 */
10 public void setBookedStateOfCar(CarDTO car,
11 boolean bookedState) {
12 CarData carToBook = findCarByRegNo(car);
13 carToBook.booked = bookedState;
14 }

Listing 8.3 The bookCar method in CarRegistry

Always use exceptions for error handling

It is never appropriate to report an error with a return value, at least not when using a program-
ming language that includes some kind of exception handling. One reason is that it might be
difficult to find appropriate return values. Consider a method that calculates xy , it might return
any number. Therefore, whatever number is assigned to indicate errors, there is the risk that it
is confused with a correct result. Another reason is that no information about the error can be
associated with a primitive return value. Also, if return values are used, it will be necessary
to wrap each method call in an if statement and check for a return value indicating an error.
With exceptions, all error handling is instead located in the catch block, separate from the
main flow in the try block. Last, but maybe most important, using a return value to indicate
an error gives low cohesion, since we are mixing two completely different meanings in the
same variable, namely a legal return value, for example the result of calculating xy , and an
error report.
This practice says that an exception must be thrown when it is not possible to book the
specified car.

Use exceptions only for error handling

This seems obvious, why throw an exception when nothing has gone wrong? There are,
however, cases where a bad public interface makes it mandatory to handle exceptions also
in a successful execution. As an example, consider a class that reads data from an array.
It has only one method, next, which returns the first unread element. Such a class could
be implemented as in listing 8.4. It is impossible to read the contents of the array without
catching ArrayIndexOutOfBoundsException, since the reader can not tell if the end of
the array is reached. A typical client of this class would contain the code in listing 8.5. To get
rid of the problematic fact that an exception is used to indicate when all elements are read, the
public interface must be improved. One more method is added, hasNext, which tells if there
are more elements. Now Iterator looks as in figure 8.6, and a typical client as in figure 8.7.
This practice does not change our decision to throw an exception when booking a car fails.

152
 Chapter 8 Handling Failure

1 public class Iterator {

2 //It is not relevant how the array is filled.
3 private Object[] theArray;
4 private int cursor = 0;
5
6 public Object next() throws
7 ArrayIndexOutOfBoundsException {
8 return theArray[cursor++];
9 }
10 }

Listing 8.4 A class that forces clients to handle exceptions even if nothing
is wrong.

1 Iterator iter = new Iterator();

2 try {
3 while (true) {
4 iter.next();
5 }
6 } catch (ArrayIndexOutOfBoundsException done) {
7 //All elements are read
8 }

Listing 8.5 A client of the class in listing 8.4 is forced to handle excep-
tions in a successful execution.

1 public class Iterator {

2 //It is not relevant how the array is filled.
3 private Object[] theArray;
4 private int cursor = 0;
5
6 public Object next() throws
7 ArrayIndexOutOfBoundsException {
8 return theArray[cursor++];
9 }
10
11 public boolean hasNext() {
12 return cursor < theArray.length;
13 }
14 }

Listing 8.6 A better implementation of the iterator.

1 Iterator iter = new Iterator();

2 while (iter.hasNext()) {

153
 Chapter 8 Handling Failure

3 System.out.println(iter.next());
4 }

Listing 8.7 A client of the improved iterator in list-

ing 8.6.

Checked or unchecked exception?

Checked exceptions are used for business logic errors. Such errors do not indicate that the
program has crashed, but that a business rule was broken. An example is a withdraw method
in a bank account. If a user tries to withdraw more money than the current balance of the
account, the method might throw an exception to indicate that the withdrawal is not allowed.
That does not mean anything is wrong with the program. Unchecked exceptions, on the other
hand, are used for programming errors. A typical example is NullPointerException,
which normally means that a method is called on an object that does not exist. This is an
indication that there is a bug in the program. Ideally, such exceptions should never occur,
except during development and testing. There is also a third category of errors, that are neither
bugs, nor business rules. These are errors that stop the program from performing its task, but
can not be eliminated during development. A typical example is if the program must call
some server to perform its work, and the server is not responding. There is no consensus on
which type of exception to use in this case, but it is probably more common to use unchecked
exceptions. A reason often mentioned is to avoid cluttering the code with catch blocks that
can not do much more than report the error.
This practice says that a checked exception shall be used to indicate that the car is already
booked. That is because booking a car that is booked would break the business rule saying a
booked car can not be booked again. The other failure reason, that something goes wrong in
the underlying storage, belongs to the category for which there is no clear advice. Let’s decide
to use an unchecked exception in this situation, since it is unlikely that anything can be done
in response to such an exception, except to report it.

Name the exception after the error condition

An exception class shall have a name that explains what went wrong, in order to clarify why
it was thrown. It is also a convention that the name ends with Exception. A good example
is the class java.lang.ArrayIndexOutOfBoundsException, whose name clearly indi-
cates that the program tried to access an element outside an array. This practice might seem
to indicate that exception classes should have names explaining in detail which situation the
class represents. That is, however, often not the best solution. Very detailed names of excep-
tion classes means many different classes will be needed, which in turn makes it mandatory
to mention all these classes in catch blocks. Also, more exception classes means a bigger
public interface, and thereby a bigger risk of having to change something in it. Therefore,
it might be better to have more generic exception classes, and instead provide detailed infor-
mation about the particular condition in the exception message. Thus, the best solution for
exception granularity is a compromise between clear information and low number of classes.

154
 Chapter 8 Handling Failure

It is now time to decide class names for the exceptions used in the book car system
operation. First, consider the checked exception thrown when the car is already booked.
It could be called something like AlreadyBookedException, but that might be too de-
tailed. At the other end of the spectrum is RentalException, which could be used for
practically any exception in the entire car rental application. A possible compromise is
CarRegistryException, which could be used for all exceptions thrown by CarRegistry,
but not by any other class. When making this decision it is important to understand that what-
ever arrives at the user interface, must inform the cashier that the car was already booked.
Otherwise, the cashier will have to tell the customer Sorry, I can not book your car and I have
no clue why. This information can only have two forms, an error code contained in the ex-
ception object, or the name of the exception class. Each form has its drawbacks. Using error
codes brings the risk of messy if statements to decide what went wrong, using class names
brings the risk of both increasing the car registry’s public interface and requiring many catch
statements to decide what went wrong. It is not obvious which option to choose, but let’s
settle for the class name. Partly because it is not very object-oriented to use if statements to
check the primitive value of the error code, and partly because, at least this far, we have not
introduced a lot of different exception classes. Later, if too many exception classes appear, we
might have to change to another solution. The outcome of this discussion is thus to call the
class AlreadyBookedException. Next, consider the case that something goes wrong in the
underlying datastore. Here, CarRegistryException is a good name. There is no detailed
information to convey, except perhaps an error code from the database API, which anyway is
of no use in the view. We are now ready to create the exception classes, see listings 8.8 and
8.9.

1 /**
2 * Thrown when trying to book a car which is already booked.
3 */
4 public class AlreadyBookedException extends Exception {
5 }

Listing 8.8 The AlreadyBookedException is created.

1 /**
2 * Thrown when something goes wrong while performing an
3 * operation in the <code>CarRegistry</code>. The message
4 * might contain more information about the error condition.
5 */
6 public class CarRegistryException extends RuntimeException {
7 }

Listing 8.9 The CarRegistryException is created.

155
 Chapter 8 Handling Failure

Include information about the error condition

It is useful, both for debugging purposes and for generating error messages, that an object
catching an exception can get information about what caused the error. Therefore, such infor-
mation shall be included in the exception object. Remember that an exception is an ordinary
object, that can have fields storing any kind of information. There are two kinds of information
that shall be included in the exception. First, all data that together make up the condition that
caused the exception. Second, a string containing an error message. This message is not in-
tended for the end user. If a message created in the model (or integration) layer is displayed in
the user interface, it effectively means that layer contains part of the view, which is definitely
not appropriate. Instead, the message is intended for developers or administrators who want
to understand what actually happened.
Regarding AlreadyBookedException, the exact error condition is known, and detailed
information can be provided. The relevant data is the car that could not be booked, which
could be stored as a CarDTO instance. The error message should state that this car was al-
ready booked, it could be Unable to book <registration number>, since it is already booked.
CarRegistryException, on the other hand, can be thrown for a number of different fail-
ures in the underlying datastore. It is therefore impossible to decide now which data to include
or what the message shall be.

Use functionality provided in java.lang.Exception

All exceptions must inherit java.lang.Exception (consult section 1.5 if you need to repeat
inheritance). In doing that, a lot of useful functionality is inherited, for example handling an
error message. The Exception class has both a constructor that takes a message, and a
method getMessage that returns the message.
Knowing the above, it is very easy to add message handling to the exception classes. All
that is needed is to add a constructor that takes the message and passes it to the superclass’
constructor. There is no need to write a get method to retrieve the message, since that method
is inherited from Exception. After this update, the classes look like in figures 8.10 and 8.11.
Note that the message is handled differently in these classes. In CarRegistryException,
since nothing is known about the message, there is nothing to do except to store it in the object.
AlreadyBookedException, on the other hand, represents only one particular condition, and
the message will always be the same. Therefore, it can be encapsulated in the exception class.
There is no reason to make it customizable, since that would only force classes throwing the
exception to bother about the message, which would mean extra error handling code in those
classes, which in turn would give them lower cohesion. The only thing that differs between
different instances of AlreadyBookedException is which car was being booked. This is
specified in the carThatCanNotBeBooked parameter in the constructor.

1 /**
2 * Thrown when trying to book a car which is already booked.
3 */
4 public class AlreadyBookedException extends Exception {

156
 Chapter 8 Handling Failure

5 private CarDTO carThatCanNotBeBooked;

6
7 /**
8 * Creates a new instance with a message specifying for
9 * which car the booking failed.
10 *
11 * @param carThatCanNotBeBooked The car that could not be
12 * booked.
13 */
14 public AlreadyBookedException(
15 CarDTO carThatCanNotBeBooked) {
16 super("Unable to book " +
17 carThatCanNotBeBooked.getRegNo() +
18 ", since it is already booked.");
19 this.carThatCanNotBeBooked = carThatCanNotBeBooked;
20 }
21
22 /**
23 * @return The car that could not be booked.
24 */
25 public CarDTO getCarThatCanNotBeBooked() {
26 return carThatCanNotBeBooked;
27 }
28
29 }

Listing 8.10 Message handling is added to AlreadyBookedException.

1 /**
2 * Thrown when something goes wrong while performing an
3 * operation in the <code>CarRegistry</code>. The message
4 * might contain more information about the error condition.
5 */
6 public class CarRegistryException extends RuntimeException {
7 /**
8 * Creates a new instance representing the condition
9 * described in the specified message.
10 *
11 * @param msg A message that describes what went wrong.
12 */
13 public CarRegistryException(String msg) {
14 super(msg);
15 }
16 }

Listing 8.11 Message handling is added to CarRegistryException.

157
 Chapter 8 Handling Failure

Use the correct abstraction level for exceptions

Say that a low-level layer (far from the view) throws some exception with a detailed technical
description of an error condition. An example could be that the integration layer fails to per-
form a database operation. Should this information be displayed to the user? Most certainly
not. First, it is not user friendly at all. What would you think if a cash machine said “sql
syntax error” when you tried to withdraw money? It would not be possible to know if the
amount had been withdrawn from the bank account, and it would probably not feel safe to use
that bank any more. Second, it could create security problems if detailed information about
the database was displayed to all users, including possible attackers. Now that it has been
established that this exception is not of interest to the user, the next question is if it then is of
any interest to the view layer? Most likely not, since anyway it shall not be displayed to the
user. Also, catching a database exception in the view creates a dependency from the view to
the integration layer, which is not desired. The conclusion is that it is often inappropriate that
an exception traverses many layers, and better to do as in listing 8.12, where the exception
from the database call is caught, and a more generic exception is thrown instead. Note that
the SQLException object, sqle, is included in the thrown OperationFailedException
object (line 15). That makes it possible for higher layers to retrieve the original
SQLException object if necessary, by calling the method getCause. This is done
on line 6 in listing 8.13. Listing 8.14 shows the OperationFailedException class.
However, note that it is not a requirement to catch an exception in each layer. Sometimes
an exception does make sense also to the next higher layer. In that case it is best to let the !
exception propagate up without catching it.
1 /**
2 * Stores the specified customer object in
3 * persistent storage.
4 * @param cust The customer to store.
5 * @throws OperationFailedException If failed to
6 * store customer.
7 */
8 public void createCustomer(Customer cust)
9 throws OperationFailedException {
10 try {
11 // Call the database.
12 } catch (SQLException sqle) {
13 throw new OperationFailedException(
14 "Could not update customer " + cust, sqle);
15 }
16 }

Listing 8.12 An exception that is too detailed for higher layers is caught, and a
more appropriate exception is thrown instead.

1 public void aMethodThatMustStoreACustomer() {

2 try {

158
 Chapter 8 Handling Failure

3 storage.createCustomer(customer);
4 } catch (OperationFailedException exc) {
5 exc.getCause(); // Returns the original exception.
6 }
7 }

Listing 8.13 How to use the method getCause().

1 public class OperationFailedException

2 extends Exception {
3
4 public OperationFailedException(String msg,
5 Exception cause) {
6 super(msg, cause);
7 }
8 }

Listing 8.14 The class OperationFailedException.

We now have to decide if CarRegistryException is appropriate for all layers, or

if it shall be wrapped in a more generic exception at some point. The exception is
thrown in the integration layer, to the method setRentedCar in Rental, in the model.
That method, in turn, is called by bookCar in the controller. Does it make sense
to let a CarRegistryException propagate up to the controller? Maybe yes, since
the controller anyway uses the CarRegistry class. That means no new dependency
is introduced if CarRegistryException continues up to the controller. On the other
hand, the code in bookCar in Controller currently does not depend on the fact that
setRentedCar calls CarRegistry. Now, if a call to setRentedCar requires catching
CarRegistryException, it also means that code must be changed if later Rental is
changed, to not call CarRegistry any more. It is not easy to decide if this problem is
big enough to motivate the introduction of a new exception class, that should be thrown by
Rental. Let’s choose the simplest solution, and not introduce a new exception, but instead let
CarRegistryException propagate up to the controller. If later there really is the need to
change the controller because of a change in setRentedCar in Rental, we might reconsider
this decision. Thus, the changes to setRentedCar are to add the throws clause on line eleven
in listing 8.15, to change the comment to clarify that this method books the car, and to check
if the car is already booked. The private method bookCar is added, in order to clarify that it is
here that the car is booked, and to give higher cohesion to setRentedCar. In fact, it might be
bad design to have a set method perform the business logic of booking the car. A set method is
normally expected to just store the specified value, and nothing more. Let’s change the method
name to rentCar, to avoid this confusion. Also, as can be seen on line 17, CarDTO has got a
new method, isBooked, which tells if the car is booked. Furthermore, a new CarRegistry
method, called getCarByRegNo, is called on line 16. It returns a CarDTO instance with the
current state of the car whose registration number is specified in the CarDTO parameter.

159
 Chapter 8 Handling Failure

1 /**
2 * Specifies the car that was rented. The specified car is
3 * also booked, thus becoming unavailable to other rentals.
4 *
5 * @param rentedCar The car that was rented.
6 */
7 public void rentCar(CarDTO rentedCar)
8 throws AlreadyBookedException {
9 bookCar(rentedCar);
10 this.rentedCar = rentedCar;
11 }
12
13 private void bookCar(CarDTO carToBook)
14 throws AlreadyBookedException {
15 CarDTO currentCarState =
16 carRegistry.getCarByRegNo(carToBook);
17 if (currentCarState.isBooked()) {
18 throw new AlreadyBookedException(currentCarState);
19 }
20 carRegistry.setBookedStateOfCar(carToBook, true);
21 }

Listing 8.15 The setRentedCar method in Rental does not catch the
CarRegistryException.

We are not done yet. The CarRegistryException is now in bookCar in the controller,
and there the same question again arises. Shall the exception be caught, or again allowed to
propagate upwards, this time to the view? Now, the answer is a definite “no”. It is never a
good design to let the view depend on the integration layer. Therefore, the exception is caught
in the controller. Still, the view must be informed of the fact that the car could not be booked.
That means a new exception must be thrown by the Controller method bookCar. This
exception can be very generic, in order to be appropriate whenever any operation fails, but
the exact reason is not of interest to the view. Let’s use the OperationFailedException,
already introduced in listing 8.14. The Controller method bookCar is shown in listing
8.16.

1 /**
2 * Books the specified car. After calling this method, the car
3 * can not be booked by any other customer. This method also
4 * permanently saves information about the current rental.
5 *
6 * @param car The car that will be booked.
7 */
8 public void bookCar(CarDTO car) throws AlreadyBookedException,
9 OperationFailedException {
10 try {

160
 Chapter 8 Handling Failure

11 rental.rentCar(car);
12 rentalRegistry.saveRental(rental);
13 } catch(CarRegistryException carRegExc) {
14 throw new OperationFailedException(
15 "Could not rent the car.",
16 carRegExc);
17 }
18 }

Listing 8.16 The bookCar method in Controller, when exception handling has been added.

This long discussion only treated CarRegistryException. We have not yet started to
consider AlreadyBookedException, but luckily, that is quite easy. Since the very reason
it was created was to convey information to the view, it most certainly shall propagate all the
way up to the view. As can be seen in listing 8.16, it is not caught in the controller, but is
instead specified in the throws clause, on line nine.

Write Javadoc comments for all exceptions

Whenever you write a method that throws an exception, also add a Javadoc comment that
describes the condition in which the exception is thrown. Such a comment is of great use
for someone trying to understand a scenario, in which the exception was thrown. One more
advantage of writing a comment, is that it forces the author to really decide what the method
shall do. The Javadoc tag for documenting an exception is @throws.
Turning to the book car case study, there are by now many methods that throw excep-
tions. Starting from the lowest layer, we first consider the methods in CarRegistry.
There are two methods in that class which throw CarRegistryException, namely
setBookedStateOfCar and getCarByRegNo. They throw the exception if the call to the
datastore fails, which will never happen since there is no datastore, or if the car can not be
found. One might ask why the exception is thrown when the car can not be found, is that
really a failure? There is no unarguable answer to this question. The reason for letting them
throw the exception is that they are supposed to perform some operation on an existing car,
not to check if that car exists. On the other hand, findAvailableCar, which searches for
a car matching a certain description, does not throw an exception if there is no matching car.
This method is not supposed to handle an existing car, but instead check if there is a certain
car. Therefore, it is not a failure if the car does not exist, but instead a valid outcome of
the method. The javadoc comments of setBookedStateOfCar and getCarByRegNo are
found in listing 8.17. The @throws tags are located on lines 9 and 25. Note that there is no
throws clause in the method declarations, since CarRegistryException is an unchecked
exception.

1 /**
2 * If there is an existing car with the registration number of
3 * the specified car, set its booked property to the specified
4 * value. Nothing is changed if the car’s booked property

161
 Chapter 8 Handling Failure

5 * already had the specified value

6 *
7 * @param car The car that shall be marked as booked.
8 * @param bookedState The new value of the booked property.
9 * @throws CarRegistryException if the database call failed,
10 * or if the specified car did
11 * not exist.
12 */
13 public void setBookedStateOfCar(CarDTO car,
14 boolean bookedState) {
15 ...
16 }
17
18 /**
19 * Searches for a car with the registration number of the
20 * specified car.
21 *
22 * @param searchedCar Searches for a car with registration
23 * number of this object.
24 * @return A car with the specified registration number.
25 * @throws CarRegistryException if the database call failed,
26 * or if the specified car did
27 * not exist.
28 */
29 public CarDTO getCarByRegNo(CarDTO searchedCar) {
30 ...
31 }

Listing 8.17 The Javadoc comments for the methods setBookedStateOfCar and
getCarByRegNo.

Continuing upwards through the layers, the next class is Rental, and the method is
rentCar. This method does not catch a CarRegistryException coming from the car
registry. Therefore, the same exception is thrown also by this method, and must therefore be
documented. Also AlreadyBookedException is included in the Javadoc comment (listing
8.18), since it is thrown by this method.

1 /**
2 * Specifies the car that was rented. The specified car is
3 * also booked, thus becoming unavailable to other rentals.
4 *
5 * @param rentedCar The car that was rented.
6 * @throws AlreadyBookedException if the car was already
7 booked.
8 * @throws CarRegistryException if the database call failed.
9 */
10 public void rentCar(CarDTO rentedCar)

162
 Chapter 8 Handling Failure

11 throws AlreadyBookedException {
12 ...
13 }

Listing 8.18 The Javadoc comment for the method rentCar.

The last method to document is bookCar in the controller. AlreadyBookedException

just passes through this method, and its comment is identical to that in rentCar, since it
makes perfect sense also here. CarRegistryException is not thrown by this method, but is
wrapped in an OperationFailedException, which tells that the book car operation could
not be performed, but the reason is unknown. The resulting Javadoc comment is in listing
8.19.

1 /**
2 * Books the specified car. After calling this method, the
3 * car can not be booked by any other customer. This method
4 * also permanently saves information about the current
5 * rental.
6 *
7 * @param car The car that will be booked.
8 * @throws AlreadyBookedException if the car was already
9 * booked.
10 * @throws OperationFailedException if unable to rent the car
11 * for any other reason than
12 * it being already booked.
13 */
14 public void bookCar(CarDTO car) throws AlreadyBookedException,
15 OperationFailedException {
16 ...
17 }

Listing 8.19 The Javadoc comment for the method bookCar.

An object shall not change state if it throws an exception

It is desirable that an object can still be used after it has thrown an exception. It would be quite
complicated if it was necessary to discard and recreate all objects involved in each operation
that resulted in an exception. If an object shall be usable, it must be clear which state it is
in, that is, which are the values of all its data elements. Throwing an exception means the
method failed, and could not perform its task. Thus, the only reasonable state of an object
after a method has thrown an exception is the same state as the object had before the method
began. Therefore, no field shall change value if an exception is thrown. There are many ways
to reach this goal, the most common follow below, sorted more or less in the order they shall
be chosen, with the most preferred strategy first.

163
 Chapter 8 Handling Failure

Make the object immutable An immutable object is an object that can never
change state. It is a programmers best friend, since there is never any doubt
about its state, and no risk that it is accidentally changed. To make an ob-
ject immutable, all of its fields must be final. Also, to be really rigid, it
must not be possible to inherit the class, and change its immutable behavior.
Listing 8.20 contains an immutable example, Person. Note that the object
address can not just be stored on line 15 in the constructor. It is necessary
to make a copy of it, otherwise also the object that created the Person would
have a reference to the same Address object, and be able to change it.

1 /**
2 * Objects of this class are immutable, none of
3 * the fields can ever change state.
4 */
5 public final class Person {
6 private final String name;
7 private final Address address;
8
9 /**
10 * Creates an instance with the specified
11 * name and address.
12 */
13 public Person(String name, Address address) {
14 this.name = name;
15 this.address = new Address(address);
16 }
17 }
18
19 /**
20 * Represents an address.
21 */
22 public class Address {
23 private String street;
24 // More fields.
25
26 /**
27 * Creates an instance with all fields equal
28 * to the fields of the specified address.
29 */
30 Address(Address address) {
31 this.street = address.street;
32 // Copy all other fields.
33 }
34 }

Listing 8.20 Objects of Person are immutable, their fields can never
change value.

164
 Chapter 8 Handling Failure

Check parameter values before changing any state A common cause of an ex-
ception is an illegal parameter value. Therefore, it is a very good practice to
let each method check for illegal parameter values first, before performing
its work. Consider for example the withdraw method in listing 8.21, which
throws OverdraftException when the amount to withdraw is larger than
the balance. Since the first thing it does is to check the parameter value (line
15), there is no risk that the state, which in this case is the balance, is changed
when the exception is thrown.

1 /**
2 * Represents a bank account.
3 */
4 public class Account {
5 private int balance;
6
7 /**
8 * Withdraws the specified amount from this
9 * account.
10 *
11 * @param amount The amount to withdraw.
12 */
13 public void withdraw(int amount) throws
14 OverdraftException {
15 if (amount > balance) {
16 throw new OverdraftException(balance,
17 amount);
18 }
19 balance = balance - amount;
20 }
21 }

Listing 8.21 This object checks if the amount to withdraw is illegal, before
it updates the balance.

Place operations that might fail before operations that alter the state This strat-
egy can be used if it is not possible to validate the parameters without per-
forming a part of the methods work. Consider, for example, a method that
shall insert a new element in a sorted collection of unique elements. The
best way to check if the new element is unique is to find the location where
it shall be inserted. This search will fail if the element is not unique, but the
collection will not be updated at that point.

Use a temporary copy of the state If none of the above strategies can be used,
we can save a temporary copy of the state and make sure to restore it before
throwing an exception.

165
 Chapter 8 Handling Failure

Now let’s check if all exception throwing methods in the case study follow this prac-
tice. Starting in CarRegistry, the methods to consider are getCarByRegNo and
setBookedStateOfCar. The first is not of interest, since it does not change any state at
all. The second must check if the car exists before changing its booked state. This is an exam-
ple where the strategy Place operations that might fail before operations that alter the state
is appropriate. To be able to change the state of the car, we must find it in the datastore. If
it is not found, the state can not be changed, but instead an exception is thrown. Continuing
to Rental, the method rentCar uses the strategy Check parameter values before changing
any state. The first thing the method does is to check that the specified car exists, the second
is to check its booked status. Only if the car exists and is not yet booked, is the state updated.
If it does not exist, or is already booked, an exception is thrown. Finally, in Controller,
bookCar does never change any state.

Never write an empty catch block

An empty catch block means that the exception is completely ignored, which is very seldom
a good thing. If en exception is thrown in the try block, execution will continue in the empty
catch block, where of course, nothing happens. Then execution will continue on the first
line below the catch block, as if no exception had been thrown. There will be no notification
whatsoever about the exception, which can make it very difficult to understand the program’s
behavior.
The exceptions in the case study are caught in the view, which we have not yet consid-
ered. Actually, there is no real view, but just hard coded calls to the system operations in
the controller. Still, since this fake view simulates an interaction with a real user, it can in-
clude error handling in the simulation. The call to the method that might throw an excep-
tion, namely bookCar, can be seen in the (slightly shortened) listing of the user interaction
simulation in listing 8.22. As stated above, the catch blocks shall not be empty, but what
shall happen there? Normally, there are two things to consider, information to users and in-
formation to developers and/or administrators. First, however, there are a few things worth
noting about listing 8.22. One thing is that the try block does not contain only the call to
bookCar, which throws exceptions, but also the rest of the method simulating the user in-
teraction. This is because if any line fails, there is no point to go on with the simulation.
Instead, execution immediately switches to error handling in the catch block. Next, not
only AlreadyBookedException and OperationFailedException are caught, but also
java.lang.Exception, which is the superclass of all exceptions. This way, any exception,
thrown on any line, will be caught and handled. No exception will ever leave our code. If
it did, the result would be some uncontrolled and undesired stack trace in the user interface.
Last, even though we have not yet decided what to write in the catch blocks, they are still
not empty. Even in the most badly designed and temporary program, we can at least call
printStackTrace to show what exception was thrown. Now, however, it is time to improve
error notification, that is the topic of the following two practices.

1 try {
2 CarDTO availableCar = new CarDTO(null, new Amount(1000),

166
 Chapter 8 Handling Failure

3 "medium", true, true,

4 "red", false);
5
6 foundCar = contr.searchMatchingCar(availableCar);
7 System.out.println("Result of searching for available " +
8 "car: " + foundCar);
9
10 AddressDTO address = new AddressDTO("Storgatan 2",
11 "12345", "Hemorten");
12 DrivingLicenseDTO drivingLicense =
13 new DrivingLicenseDTO("982193721937213");
14 CustomerDTO customer = new CustomerDTO("Stina", address,
15 drivingLicense);
16 contr.registerCustomer(customer);
17 System.out.println("Customer is registered");
18
19 contr.bookCar(foundCar);
20 System.out.println("Car is booked");
21 Amount paidAmount = new Amount(1500);
22 System.out.println("----- Receipt follows -----");
23 contr.pay(paidAmount);
24 System.out.println("----- End of receipt -----");
25 } catch (AlreadyBookedException exc) {
26 exc.printStackTrace();
27 } catch (OperationFailedException exc) {
28 exc.printStackTrace();
29 } catch (Exception exc) {
30 exc.printStackTrace();
31 }

Listing 8.22 The try-catch block in the view.

How to notify the user?

The user interface must always reflect the state of the system, and clearly show what is hap-
pening. If the user has initiated a system operation, and that operation can not be successfully
completed, the user interface must show this. As has been discussed previously, it is not al-
ways adequate to show very detailed error messages, for example containing database error
codes. It is, instead, necessary to carefully consider what is the most appropriate error mes-
sage, specifying what the user needs to know, without creating confusion. It is also important
that information is always given consistently. The user must not be surprised or confused by
getting different messages from different parts of the system. Messages shall always have
similar format, and two similar messages must mean the same thing. As a consequence of
this reasoning, it is a common design choice to have only one component generating error
messages. This is also the best solution in terms of cohesion and encapsulation. Do not spread

167
 Chapter 8 Handling Failure

similar error handling code over many catch blocks, but instead just call the component
responsible for error messages, and encapsulate user interface handling in there.
The case study does not have a real user interface, but we can still create a class that
prints error messages to System.out, see listing 8.23. It is called from the catch blocks,
as in listing 8.24. Since there was no need for a unique error message in answer to an
OperationFailedException, it is no longer caught explicitly, but is handled by the catch
block for any java.lang.Exception. Note that the error message is created in the view, it
is not the string returned by getMessage in Exception, since that would have meant some
lower layer decided what to print in the user interface.

1 /**
2 * This class is responsible for showing error messages to
3 * the user.
4 */
5 public class ErrorMessageHandler {
6
7 /**
8 * Displays the specified error message.
9 *
10 * @param msg The error message.
11 */
12 void showErrorMsg(String msg) {
13 StringBuilder errorMsgBuilder = new StringBuilder();
14 errorMsgBuilder.append(createTime());
15 errorMsgBuilder.append(", ERROR: ");
16 errorMsgBuilder.append(msg);
17 System.out.println(errorMsgBuilder);
18 }
19
20 private String createTime() {
21 LocalDateTime now = LocalDateTime.now();
22 DateTimeFormatter formatter = DateTimeFormatter.
23 ofLocalizedDateTime(FormatStyle.MEDIUM);
24 return now.format(formatter);
25 }
26 }

Listing 8.23 The class ErrorMessageHandler, which is responsible for showing error mes-
sages to the user.

1 } catch (AlreadyBookedException exc) {

2 errorMsgHandler.showErrorMsg(
3 exc.getCarThatCanNotBeBooked().getRegNo() +
4 " is already booked.");
5 } catch (Exception exc) {
6 errorMsgHandler.showErrorMsg("Failed to book, please" +

168
 Chapter 8 Handling Failure

7 " try again.");

8 }

Listing 8.24 The calls to ErrorMessageHandler.

How to notify developers and/or administrators?

It is not enough to inform the user about the application’s state. It must also be possible
for application administrators to understand what has been performed, and for developers to
check if there are bugs. This requires more detailed reports about what happens, for example
exception stack traces. These are printed to the log, which can be for example a text file.
There is normally one dedicated component, which is responsible for creating log entries.
The reason for creating such a component is exactly the same as for creating the component
generating user interface error messages. That is, messages shall be similar no matter why or
where they are caused, log handling shall not be duplicated in many different locations, and
detailed knowledge about the logging procedure shall be encapsulated in one component.
The case study logs to a file called rentcar-log.txt, the class handling it can be seen in
listing 8.25. There already exists logging APIs, for example the one in java.util.logging
that format messages, include time and origin, and manage log files. Normally, such an API
should be used, instead of, as is done in this listing, creating a new class performing all these
tasks. The only reason to rewrite the functionality here, instead of using an existing API, is to
keep this case study shorter, and not force the reader to learn a new API. Example calls of the
logging component can be seen in listing 8.26. Both catch blocks contain the same two lines
of code, which means these two lines are duplicated. This means a private method should be
introduced, as in listing 8.27

1 /**
2 * This class is responsible for the log.
3 */
4 public class LogHandler {
5 private static final String LOG_FILE_NAME =
6 "rentcar-log.txt";
7 private PrintWriter logFile;
8
9 public LogHandler() throws IOException {
10 logFile = new PrintWriter(
11 new FileWriter(LOG_FILE_NAME), true);
12 }
13
14 /**
15 * Writes log entries.
16 *
17 * @param entry The log entry.
18 */

169
 Chapter 8 Handling Failure

19 public void logException(Exception exception) {

20 StringBuilder logMsgBuilder = new StringBuilder();
21 logMsgBuilder.append(createTime());
22 logMsgBuilder.append(", Exception was thrown: ");
23 logMsgBuilder.append(exception.getMessage());
24 logFile.println(logMsgBuilder);
25 exception.printStackTrace(logFile);
26 }
27
28 private String createTime() {
29 LocalDateTime now = LocalDateTime.now();
30 DateTimeFormatter formatter = DateTimeFormatter.
31 ofLocalizedDateTime(FormatStyle.MEDIUM);
32 return now.format(formatter);
33 }
34 }

Listing 8.25 The class LogHandler, which is responsible for logging.

1 } catch (AlreadyBookedException exc) {

2 errorMsgHandler.showErrorMsg(
3 exc.getCarThatCanNotBeBooked().getRegNo() +
4 " is already booked.");
5 logger.logException(exc);
6 } catch (Exception exc) {
7 errorMsgHandler.showErrorMsg("Failed to book, please " +
8 " try again.");
9 logger.logException(exc);
10 }

Listing 8.26 The calls to LogHandler.

1 } catch (AlreadyBookedException exc) {

2 handleException(exc.getCarThatCanNotBeBooked().getRegNo()
3 + " is already booked.", exc);
4 } catch (Exception exc) {
5 handleException("Failed to book, please try again.", exc);
6 }
7
8 private void handleException(String uiMsg, Exception exc) {
9 errorMsgHandler.showErrorMsg(uiMsg);
10 logger.logException(exc);
11 }

Listing 8.27 The calls to LogHandler, without duplicated code.

170
 Chapter 8 Handling Failure

Write unit tests for the exception handling

Exception handling must of course be unit tested, just like everything else. Things to test
are that exceptions are not thrown during successful execution, that they are thrown when
failures occur, that messages and other content of exception objects are correct, and that they
are handled as intended in catch blocks.
Starting from the bottommost layer of the case study, the tests for exception handling in
CarRegistry are calls to getCarByRegNo and setBookedStateOfCar, with a car that
does not exist. CarRegistryException shall be thrown in both cases. Exception handling
tests for Rental and Controller are the same, to rent a car that is already booked, and to
rent a car that does not exist. Finally, there are tests also for the two classes responsible for
error messages in the user interface and in the log file. The tests for Rental can be found in
listing 8.28, as an example of exception handling unit tests. The complete tests are included
in the accompanying NetBeans project [Code].

1 @Test
2 public void testSetRentedCarWhenCarIsBooked() {
3 CarRegistry carReg = new RegistryCreator().
4 getCarRegistry();
5 Rental instance = new Rental(null, carReg);
6 CarDTO rentedCar = new CarDTO("abc123", new Amount(1000),
7 "medium", true,
8 true, "red", false);
9 try {
10 carReg.setBookedStateOfCar(rentedCar, true);
11 instance.rentCar(rentedCar);
12 fail("Could rent a booked car.");
13 } catch (AlreadyBookedException ex) {
14 assertTrue("Wrong exception message, does not " +
15 "contain specified car: " +ex.getMessage(),
16 ex.getMessage().contains(
17 rentedCar.getRegNo()));
18 assertTrue("Wrong car is specified: " +
19 ex.getCarThatCanNotBeBooked(),
20 ex.getCarThatCanNotBeBooked().getRegNo().
21 equals(rentedCar.getRegNo()));
22 }
23 }
24
25 @Test
26 public void testSetRentedCarWhenCarDoesNotExist() throws
27 AlreadyBookedException {
28 CarRegistry carReg =
29 new RegistryCreator().getCarRegistry();
30 Rental instance = new Rental(null, carReg);
31 CarDTO rentedCar = new CarDTO("wrong", new Amount(1000),

171
 Chapter 8 Handling Failure

32 "medium", true,
33 true, "red", false);
34 try {
35 instance.rentCar(rentedCar);
36 fail("Could rent a non-existing car.");
37 } catch (CarRegistryException exc) {
38 assertTrue("Wrong exception message, does not " +
39 " contain specified car: "
40 + exc.getMessage(),
41 exc.getMessage().contains(
42 rentedCar.toString()));
43 }
44 }

Listing 8.28 Unit testing the exception handling in Rental.

8.3 Complete Exception Handling in the Case Study

It is not obvious where and when to use exceptions in the rent car case study, since the code
is still quite short and incomplete. For example, should all methods in CarRegistry throw
CarRegistryException, to simulate database failures, even though there is no database
yet? In this implementation, an exception is thrown only when there are obvious failures,
in order to keep the case study reasonably short and clear. In fact, there are only two more
situations where exceptions must be used, besides the car booking described above.
The first situation requiring an exception is when LogHandler fails to open the log file.
In this case, an IOException is thrown. Since this happens during startup, the exception
is caught in main It is an open question if the program shall be allowed to continue without
logging. Currently, it is terminated if the log handler fails.
The second situation where an exception is needed, is due to the fact that the controller op-
erations registerCustomer, bookCar, and pay, must be called in that order. The reason
why this order is required is that registerCustomer creates the rental object used in all
three methods, therefore a NullPointerException will be thrown if one of the other meth-
ods is called first. Also, since each call to registerCustomer creates a new Rental, any
ongoing rental will be lost when that method is called. It can of course be questioned if this
implementation is correct, but that is really a question about business rules, which can only be
answered by a domain expert. In the current implementation, an IllegalStateException
is thrown if these operations are called in any other order. This is the exception usually used
when methods are called in wrong order.
The implementation of these scenarios, together with its unit tests, can be found in the
NetBeans project in GitHub [Code]. None of the code is listed here, since it contains nothing
conceptually new, and no particularly tricky issues.

172
 Chapter 8 Handling Failure

8.4 Common Mistakes

Even when following all the best practices mentioned in section 8.2, it is still possible to make
mistakes. Below are two common mistakes.

Failure handling is missing The fact that the program works does not necessarily mean that
failure handling is correct. It is a common mistake not handle all possible failures, for
example not to check that all parameters in a method have valid values. A good way
to discover this flaw is to write extensive unit tests. Unit tests shall check all possible
parameter values, and all possible execution paths, which means erroneous conditions
will be discovered.
NO!
Exceptions are caught when it is not needed Do not catch an exception if there is nothing
to do in the catch block. It is quite common to misunderstand the practice saying
Use the correct abstraction level for exceptions, and always catch all exceptions in
all methods. The result is code similar to listing 8.29, where, on lines four to six,
the exception is caught just to be rethrown. If the exception is appropriate also in the
calling method, then it is best to just let it propagate further up in the call stack, as in
listing 8.30.
1 public void myMethod() throws MyException {
2 try {
3 methodThatThrowsMyException();
4 catch(MyException exc) {
5 throw exc;
6 }
7 }

Listing 8.29 There is no point in catching an exception just to immediately rethrow it, as in this
listing.

1 public void myMethod() throws MyException {

2 methodThatThrowsMyException();
3 }

Listing 8.30 If the exception makes sense also in the calling method, it can continue to that method.

173
Chapter 9

Polymorphism and Inheritance

All design discussions up to this point have been based on the fundamental concepts coupling,
cohesion and encapsulation. It is now time to introduce two more concepts, polymorphism and
inheritance, and design solutions based on those. Especially polymorphism, but also inheri-
tance, are very powerful tools for creating a design. Still, the quality of these, more advanced,
designs will be completely decided by the criteria coupling, cohesion and encapsulation. The
only value of polymorphism and inheritance is to make the design meet those criteria to a
higher extent.
It is not possible to understand this chapter without a basic understanding of the concepts
interface, implementation and inheritance. Make sure you fully understand sections 1.8 !
and 1.9 before reading this chapter.

9.1 UML
In order to design polymorphism and inheritance, it is nec-
essary to model interfaces, implementations and inheri-
tances in UML. How do this in a class diagram is illustrated
in figure 9.1, where ClassA implements InterfaceA,
and ClassB inherits ClassC. There are a few sub-
tle things worth emphasizing here. First, the stereotype
«interface» is used to tell that InterfaceA in an in-
terface and not a class. A stereotype is a categorization.
Here, it says that InterfaceA belongs to the category in-
(a) terfaces. This stereotype is the only difference between the
symbols for class and interface. Another issue is that the
word implementation does not exist in UML, instead, the
same thing is called realization. Also, the word inheritance
is very vaguely specified and not much used. Instead, a su-
perclass is said to be a generalization of a subclass. Fur-
thermore, note that the method methodA in the interface is
written in italics. This means it is abstract, it consists only
(b)
of a declaration and does not have any body.
Figure 9.1 Class diagram, illustrating: It is an unfortunate fact that there is now way to il-
(a) interface and implementation lustrate implementation or inheritance in a sequence or
(b) inheritance

174
 Chapter 9 Polymorphism and Inheritance

a communication diagram. If, for example, an object of ClassA in figure 9.1a is to

be used in a sequence diagram, it must be of either the type ClassA (figure 9.2a) or
InterfaceA (figure 9.2b). It is impossible to show that the object has both types.
It is in fact a direct error to draw both types, as in figure 9.2c. That would mean there were
two different objects, one of type InterfaceA and one of type ClassA.
NO!

(a) (b)

(c)

Figure 9.2 An object of a class that implements an interface can be illustrated as the class, figure (a), or as the
interface, figure (b), but not as both, figure (c).

There are two different ways to draw an interface in

a communication diagram and a sequence diagram. It
can be drawn either as in figure 9.2, using the same
symbol as in a class diagram, or it can be drawn using
the symbol in figure 9.3.

9.2 Polymorphism
Polymorphism is Greek for many forms, the result of
Figure 9.3 Using icon notation to illustrate
an interface. using polymorphism is that one single public interface
can have any number of different implementations.
How to achieve this will be explained with the help
of a concrete example, namely a logging API. For a start, this API consists of one single class,
which can write log entries to a file, listing 9.1. The public interface of this class is colored
blue in the listing. It consists of the class name, the definition of the constructor, and the
definition of the log method. Which part of this public interface is really necessary to know

175
 Chapter 9 Polymorphism and Inheritance

for an object that wants to log some-

thing? In fact, it is only the log method.
That method must be called for a log en-
try to be written to the file, and therefore
its declaration must be known to a client
of the API. The constructor, on the other
hand, is not mandatory knowledge for a
log client that does not create an object Figure 9.4 Polymorphism in nature: light and dark morph
of FileLogger. Also the class name (form) jaguars. Image by en:User:Cburnett - Own work, CC BY-SA
3.0, https://commons.wikimedia.org/w/index.php?curid=1284082 (top),
could be hidden from a client, if it was and By Ron Singer - U.S. Fish and Wildlife Service, Public Domain,
possible to give the client a reference to https://commons.wikimedia.org/w/index.php?curid=314156 (bottom).
some object, of an unspecified class, that
has a log method. The client would then be able to use that object to call the log method,
and have the entry printed to the log file, without caring about the class name. To fully under-
stand this fact, consider listing 9.2, which contains a client of the log API. The constructor of
FileLogger is not used anywhere, and the class name, FileLogger, is used only on lines
eight and ten. To get rid of the dependency on this class name, it would suffice to remove
it from those two places. This can be achieved by introducing an interface specifying the
mandatory knowledge about the API, which, as stated above, is only the log method. Let’s
create such an interface, and call it Logger, listing 9.3. Also, FileLogger shall be changed
to implement this interface, but no other change is required to FileLogger, since it already
has the log method specified in the interface. Now, a client of the API does not need to know
anything besides the content of this interface, which is proved by listing 9.4.

1 package se.leiflindback.oodbook.polymorphism.logapi;
2
3 import java.io.FileWriter;
4 import java.io.IOException;
5 import java.io.PrintWriter;
6
7 /**
8 * Prints log messages to a file. The log file will be in the
9 * current directory and will be called log.txt.
10 */
11 public class FileLogger {
12 private PrintWriter logStream;
13
14 /**
15 * Creates a new instance and also creates a new log file.
16 * An existing log file will be deleted.
17 */
18 public FileLogger() {
19 try {
20 logStream = new PrintWriter(
21 new FileWriter("log.txt"), true);

176
 Chapter 9 Polymorphism and Inheritance

22 } catch (IOException ioe) {

23 System.out.println("CAN NOT LOG.");
24 ioe.printStackTrace();
25 }
26 }
27
28 /**
29 * Prints the specified string to the log file.
30 *
31 * @param message The string that will be printed to the
32 * log file.
33 */
34 public void log(String message) {
35 logStream.println(message);
36 }
37 }

Listing 9.1 In the first version, the logging API contains only this class. The public interface is
colored blue.

1 package se.leiflindback.oodbook.polymorphism.logapi;
2
3 /**
4 * A client for the logger. Prints log messages to the
5 * specified logger.
6 */
7 public class AnyClassThatNeedsToLogSomething {
8 private FileLogger logger;
9
10 public void setLogger(FileLogger logger) {
11 this.logger = logger;
12 }
13
14 /**
15 * Prints to the log. The logged string includes the
16 * specified message number.
17 *
18 * @param msgNo This number is included in the logged
19 * string.
20 */
21 public void anyMethod(int msgNo) {
22 logger.log("Important message number " + msgNo);
23 }
24 }

Listing 9.2 A client of the log API in listing 9.1.

177
 Chapter 9 Polymorphism and Inheritance

1 package se.leiflindback.oodbook.polymorphism.logapi;
2
3 /**
4 * Specifies an object that can print to a log. This interface
5 * does not handle log locations, it is up to the implementing
6 * class to decide where the log is.
7 */
8 public interface Logger {
9
10 /**
11 * The specified message is printed to the log.
12 *
13 * @param message The message that will be logged.
14 */
15 void log(String message);
16 }

Listing 9.3 An interface specifying the functionality "write to the log".

1 package se.leiflindback.oodbook.polymorphism.logapi;
2
3 /**
4 * A client for the logger. Prints log messages to the
5 * specified logger.
6 */
7 public class AnyClassThatNeedsToLogSomething {
8 private Logger logger;
9
10 public void setLogger(Logger logger) {
11 this.logger = logger;
12 }
13
14 /**
15 * Prints to the log. The logged string includes the
16 * specified message number.
17 *
18 * @param msgNo This number is included in the logged
19 * string.
20 */
21 public void anyMethod(int msgNo) {
22 logger.log("Important message number " + msgNo);
23 }
24 }

Listing 9.4 A client of the log API, when FileLogger implements the Logger interface
in listing 9.3.

178
 Chapter 9 Polymorphism and Inheritance

This code is quite amazing already now. It contains an API that has encapsulated the name
of the class doing the work of the API! This is extremely low coupling, the client is using
a class, FileLogger, without any dependency on the name of that class. Still, we are far
from done, it is about to become much more amazing. Now is the time to actually make
use of polymorphism, that is, to provide more than one implementation of a single public
interface. The public interface for the logging API consists of the log method in the logger
interface. This public interface currently has one implementation, provided by FileLogger.
The second implementation will be a class, ConsoleLogger in listing 9.5, which prints the
log messages to the screen, instead of to a file. Also this implementation fulfills the contract
of the public interface defined by log, which, according to the JavaDoc, is that The specified
message is printed to the log.

1 package se.leiflindback.oodbook.polymorphism.logapi;
2
3 /**
4 * Prints log messages to <code>System.out</code>.
5 */
6 public class ConsoleLogger implements Logger {
7
8 /**
9 * Prints the specified string to <code>System.out</code>.
10 *
11 * @param message The string that will be printed to
12 * <code>System.out</code>.
13 */
14 @Override
15 public void log(String message) {
16 System.out.println(message);
17 }
18 }

Listing 9.5 The second implementation of the Logger interface is this class, ConsoleLogger,
which prints the log messages to the screen.

Figure 9.5 The log API and a class using it.

179
 Chapter 9 Polymorphism and Inheritance

Figure 9.5 contains a class diagram with all classes created so far. The client of the log API,
AnyClassThatNeedsToLogSomething depends only on the interface Logger. It has no
coupling whatsoever to the implementations of the interface! This fact is the foundation of all
fancy usages of polymorphism. It means that AnyClassThatNeedsToLogSomething does
not know, and has no interest in, which class it is actually calling. All that matters is that it is a
class providing an implementation of the required public interface. This is guaranteed by the
compiler, since any object passed to setLogger must be of a class implementing Logger,
and must therefore provide an implementation of the public interface defined in Logger. It is
now possible to change implementation at runtime, as is done when using the main method
of the class Main, in listing 9.6. Make sure you fully understand this example, similar designs
are extremely common in object-oriented programming.

1 package se.leiflindback.oodbook.polymorphism.logapi;
2
3 /**
4 * Contains the main method of the log API client.
5 */
6 public class Main {
7
8 /**
9 * @param args The program does not take any command line
10 * parameters.
11 */
12 public static void main(String[] args) {
13 AnyClassThatNeedsToLogSomething client
14 = new AnyClassThatNeedsToLogSomething();
15
16 client.setLogger(new FileLogger());
17 client.anyMethod(1);
18 client.anyMethod(2);
19 client.anyMethod(3);
20
21 client.setLogger(new ConsoleLogger());
22 client.anyMethod(4);
23 client.anyMethod(5);
24 client.anyMethod(6);
25 }
26 }

Listing 9.6 A main method of a program using the log API. The first three logged messages are
printed to the file, the last three to the screen.

In conclusion, by using polymorphism the program gets higher cohesion, since the client
of the polymorphic design, AnyClassThatNeedsToLogSomething in the example above,
does not contain any code related to choosing implementation of the public interface. It just
uses the implementation passed to it (in the method setLogger). There will also be lower

180
 Chapter 9 Polymorphism and Inheritance

coupling, since there is no coupling from the client to any implementation. Finally, the pro-
gram gets better encapsulation, since the names of the implementing classes are completely
encapsulated inside the logging API.
The result of improving both coupling, cohesion and encapsulation, is that the behavior of
the program can be changed at runtime, by switching implementation of a public interface.
This was done in the above example by changing from FileLogger to ConsoleLogger.
Without polymorphism, the destination of the logs would have been hard-coded in the pro-
gram. The only way to change logging at runtime would have been with if statements, like
the code in listing 9.7. Such if statements would have had to be repeated in every class that
wanted to log something. Also, they would have to be updated whenever a new implementa-
tion was added.

1 if (logToFile) {
2 Write to the file
3 } else if(logToScreen) {
4 Write to the screen
5 } similar statements checking other log destinations

Listing 9.7 Pseudocode showing if statements replacing polymorphism.

Before leaving this introduction to polymorphism, as an extra exercise let’s look at an

example where implementations are not only changed at runtime, but also added at run-
time. Many programming languages, including Java, have the possibility to load and
execute new code, that did not exist when the program started. This is illustrated in
the main method in listing 9.8. The actual loading and instantiation of a new class
is done on lines 32 and 33. The fully qualified class names of Logger implemen-
tations that shall be used, shall be command line parameters, separated by a space.
To make the class use the FileLogger and ConsoleLogger, the command line pa-
rameters shall be se.leiflindback.oodbook.polymorphism.logapi.FileLogger
se.leiflindback.oodbook.polymorphism.logapi.ConsoleLogger. Note the space
between the class names.

1 package se.leiflindback.oodbook.polymorphism.logapi;
2
3 /**
4 * Contains a main method of the log API client, which loads
5 * new <code>Logger</code> implementations at runtime.
6 */
7 public class LoadImplAtRuntime {
8 private int msgNo = 1;
9 private AnyClassThatNeedsToLogSomething client =
10 new AnyClassThatNeedsToLogSomething();
11
12 /**
13 * @param args Each command line parameter shall be the

181
 Chapter 9 Polymorphism and Inheritance

14 * fully qualified class name of a class

15 * implementing <code>Logger</code>. This
16 * class will be loaded and used.
17 */
18 public static void main(String[] args) throws
19 InstantiationException,
20 IllegalAccessException,
21 ClassNotFoundException {
22 LoadImplAtRuntime main = new LoadImplAtRuntime();
23 for (String logger : args) {
24 main.loadAndUseLogger(logger);
25 }
26 }
27
28 private void loadAndUseLogger(String logger) throws
29 InstantiationException,
30 IllegalAccessException,
31 ClassNotFoundException {
32 Class logClass = Class.forName(logger);
33 Logger logInstance = (Logger)logClass.newInstance();
34 client.setLogger(logInstance);
35 client.anyMethod(msgNo++);
36 }
37 }

Listing 9.8 A main method that can load classes at runtime. To make this class load and
use FileLogger (listing 9.1) and ConsoleLogger (listing 9.5), the command line pa-
rameters shall be se.leiflindback.oodbook.polymorphism.logapi.FileLogger
se.leiflindback.oodbook.polymorphism.logapi.ConsoleLogger

9.3 Inheritance
Inheritance is a very special kind of relation between classes. If one class, the subclass, inherits
another class, the superclass, it means that all non-private members of the superclass also
become members of the subclass. In fact, all code in the superclass, except code with private
visibility, becomes code also in the subclass. This very special kind of relation has severe
consequences, as will be explained below. Before proceeding, it might adequate to repeat the
basics of inheritance, by reading section 1.9.
Another thing that must be understood is protected visibility. This is the fourth type of
visibility, besides public, private and package private. A member with protected visibility can
be accessed by a subclass in any package, but by a non-subclass only in the same package. This
is illustrated in figure 9.6. Note that the symbol for protected visibility is the hash character,
#. Since the protected member is visible to classes in any package, it is part of the public
interface, and not of the implementation. It is thus more closely related to public visibility
than to private or package private.

182
 Chapter 9 Polymorphism and Inheritance

(a) (b)

(c) (d)

Figure 9.6 The method protectedMethod, which has protected visibility, can be called by a method in
Class2 in all situations except diagram (d).

Why Not to Use Inheritance, And What to do Instead

Never use inheritance just to reuse code from another class. Code reuse is much better
achieved by simply calling the methods that shall be reused, instead of inheriting them from a
super class. This strategy, to use a plain, ordinary association instead of inheritance, is called
composition. The reasons to prefer composition are explained below.

First Reason to Prefer Composition: Inheritance Complicates Code Reuse

Figure 9.7 An example of a classical inheritance hierarchy, representing different species.

Much is written on creating classical inheritance hierarchies, like the one in figure 9.7.
However, they are often not as useful as it might seem, but instead make it difficult to reuse
code. This class hierarchy describes animals, now let’s focus on how they move. The diagram

183
 Chapter 9 Polymorphism and Inheritance

shows that birds fly, fishes swim and mammals walk. Seems OK, but what if want to add a
penguin, which is a bird that swims but does not fly? Or a mammal that can both swim and
walk? And what about a creature that changes behavior as it grows, for example a bird that can
not fly when it is hatched, but later learns to? This reasoning might suggest it would be better
to create a class hierarchy based on movement possibilities than on taxonomy, as in figure 9.8.

Figure 9.8 The animal inheritance hierarchy, based on how species move, instead of taxonomy.

Unfortunately, this change does not solve the problem. First, we would need multiple in-
heritance for the mammal that both swims and walks. Second, it would still be impossible to
change movement for the bird that learns to fly. Third, there are many properties of animals
which do not fit in the movement hierarchy. Consider for example wings, ostriches have wings
but can not fly, so the wing property can not be placed in Flyer. Above all, why create any
hierarchy at all? Why not use composition instead, as in figure 9.9? This way, any animal
object can be connected to any movement object, or to several movement objects. Also, an
animal object can, at any point in time, change the set of available movement objects. The
solution should be improved further, using polymorphism for the movement, and maybe the
strategy pattern described below. However, what is important here is only the advantage of
composition over inheritance. To be honest, though, it must be admitted that the diagram in
figure 9.9 is messier than those in figures 9.7 and 9.8. This is the disadvantage of composition,
that it becomes more difficult to understand which class has a reference to which, especially
when references change at runtime, and even more so if polymorphism is used.

Second Reason to Prefer Composition: Inheritance Breaks Encapsulation

Inheritance makes the code difficult to maintain, and may introduce bugs, since it breaks
encapsulation. Not only the public interface of the superclass is inherited by the subclass, but
also the implementation. Simply speaking, everything in the superclass becomes a part also
of the subclass. It should be obvious that the mere fact of completely breaking encapsulation
is something bad. Now, let’s look at an example illustrating the risk of inheriting a class that
was not written to be inherited, just for the purpose of code reuse.
Suppose we have at our disposal an API, that includes a class List, part of which is shown
in listing 9.9. Unfortunately, this class does not really meet our needs, since we want a list

184
 Chapter 9 Polymorphism and Inheritance

Figure 9.9 Using composition, instead of inheritance, for the animal classes.

that counts how many items have been added during the entire existence of a List object (not
how many items there are currently in the list). We therefore decide to create a new class,
CountingList, which inherits List and adds the missing functionality. This can be done as
in listing 9.10.

1 public class List {

2 private static final int INITIAL_SIZE = 100;
3 private Object[] content = new Object[INITIAL_SIZE];
4 private int firstFreeIndex = 0;
5
6 /**
7 * Adds an element last in the list.
8 * @param element The element to add.
9 */
10 public void add(Object element) {
11 content[firstFreeIndex++] = element;
12 }
13
14 /**
15 * Adds all elements in the specified list to this list.
16 * @param elemsToAdd The elements to add.
17 */
18 public void addAll(List elemsToAdd) {
19 // Add all elements in elemsToAdd.
20 }
21 }

Listing 9.9 Part of a list implementation.

185
 Chapter 9 Polymorphism and Inheritance

1 public class CountingList extends List {

2 private int noOfAddedElems;
3
4 public void add(Object elemToAdd) {
5 noOfAddedElems++;
6 super.add(elemToAdd);
7 }
8
9 public void addAll(List elemsToAdd) {
10 noOfAddedElems = noOfAddedElems + elemsToAdd.size();
11 super.addAll(elemsToAdd);
12 }
13 }

Listing 9.10 An attempt to extend the list from listing 9.9, and count how many items have ever
been added to the list.

This seems to be a good solution, but it does not work. When addAll is called, the amount
of added elements is doubled. Adding five elements increases the noOfAddedElems counter
by ten. The reason is the addAll in List was implemented as in listing 9.11. This method
iterates over the list and calls add for each element, in order not to duplicate the code that
adds a single element. However, that call is now to add in CountingList, and thereby the
counter is incremented twice, both in addAll and add.

1 public void addAll(List elemsToAdd) {

2 for (int i=0; i<elemsToAdd.size(); i++) {
3 add(elemsToAdd.get(i));
4 }
5 }

Listing 9.11 The implemenation of addAll. The call to add on

line three will be to add in CountingList.

There are of course different ways to get rid of this bug, for example not to increment
noOfAddedElems in addAll. However, such solutions only make CountingList ever
more dependent on particularities of List’s implementation. This implementation is of course
not documented, since it is supposed to be encapsulated. Also, it might very well be that the
List implementation changes between releases, without any notice, which might introduce
new bugs in CountingList. The conclusion is that inheritance is not appropriate here. It
is much better to reuse the existing list class with composition, as in listing 9.12. Just as
in the animal example above, composition makes the code a bit longer. In this case since
CountingList must include all List methods that shall be available. It is, in fact, not un-
common that composition is less elegant than inheritance seems to be, at first sight. However,
composition has the big advantage that it actually works, and it is much less likely to create
problems in the future, when the code is changed.

186
 Chapter 9 Polymorphism and Inheritance

1 /**
2 * A list that counts how many elements have ever been added.
3 */
4 public class CountingListUsingComposition {
5 private int noOfAddedElems;
6 private List list = new List();
7
8 public void add(Object elemToAdd) {
9 noOfAddedElems++;
10 list.add(elemToAdd);
11 }
12
13 public void addAll(
14 CountingListUsingComposition elemsToAdd) {
15 noOfAddedElems = noOfAddedElems + elemsToAdd.size();
16 list.addAll(elemsToAdd.list);
17 }
18
19 /**
20 * Tells how many elements have ever been added to
21 * the list.
22 *
23 * @return the number of elements that have ever been
24 * added to the list.
25 */
26 public int noOfAddedElems() {
27 return noOfAddedElems;
28 }
29
30 public int size() {
31 return list.size();
32 }
33
34 public Object get(int index) {
35 return list.get(index);
36 }
37 }

Listing 9.12 Code reuse with composition, instead of inheritance.

Third Reason to Prefer Composition: There is no Way to Control the Public Interface of
the Subclass

A subclass blindly accepts the entire public interface of its superclass. If those classes have
different authors, the subclass’ author has effectively given up control of the created class’

187
 Chapter 9 Polymorphism and Inheritance

public interface. Not a very nice situation. This problem arises when using inheritance for
code reuse, thus inheriting a class whose author does not know it has been inherited. This
is the situation with the code in listing 9.10, where CountingList has the entire public
interface of List. In that case, this might seem appropriate, but can we really be sure we want
everything from List to appear also in CountingList? As an example, java.util.List
has 35 methods (in JDK 8). If the list class inherited here is of a similar size, it is quite an
effort just to go through all those methods and decide if they suit also the subclass. Then,
it is necessary to keep track of when new versions of the superclass are released, and check
those for changes to the public interface. If a new superclass version is not checked, the public
interface of the subclass might change without anyone changing its code, and without anyone
knowing it changed.

How to Use Inheritance Appropriately

Given the drawbacks discussed above, it might seem that inheritance should never be used at
all. This is of course not the case. Used correctly, inheritance can give really beautiful code
that is easy to understand and maintain. Below are typical situations where inheritance should
be used.

Use Inheritance to Modify Behavior

It is sometimes the case that the task of one class is very similar to the task of another class,
except in some detail. Consider for example a class that creates a bar chart illustrating a data
set, which is read from a database. Such a class could be written as in listing 9.13

1 public class BarChart {

2 private String dataId;
3
4 public Chart(String dataId) {
5 this.dataId = dataId;
6 }
7
8 public Image getChart() {
9 try {
10 // read the specified data from database.
11 // create image with bar chart.
12 return barChartImage;
13 } catch(SQLException sqle) {
14 // exception handling
15 }
16 return null;
17 }
18 }

Listing 9.13 Pseudocode for a class that reads data from a database, and presents it in a bar chart.

188
 Chapter 9 Polymorphism and Inheritance

Now suppose we also want to present the data as a line chart. This would require another
class, LineChart, which would be identical to BarChart except for the image creation on
line eleven. Obviously there is duplicated code, but how to remove it? the problem is that
the duplicated code forms the structure of the class, the handling of dataId, the try-catch
block, etc. A solution is to create a superclass that contains all common code, and place the
specific code in a method that can be overridden in subclasses, as in listing 9.14.
This way of placing implementation specific code in a protected method is in fact a design
pattern called template method, which is explained in more detail below, in section 9.4.

1 public abstract class Chart {

2 private String dataId;
3
4 public Chart(String dataId) {
5 this.dataId = dataId;
6 }
7
8 public Image getChart() {
9 try {
10 // read the specified data from database.
11 Image chartImage = createChartImage();
12 return chartImage;
13 } catch(SQLException sqle) {
14 // exception handling
15 }
16 return null;
17 }
18
19 protected abstract Image createChartImage();
20 }
21
22 public class BarChart extends Chart {
23 public Chart(String dataId) {
24 super(dataId);
25 }
26
27 protected Image createChartImage() {
28 // create and return bar chart.
29 }
30 }
31
32 public class LineChart extends Chart {
33 public Chart(String dataId) {
34 super(dataId);
35 }
36
37 protected Image createChartImage() {

189
 Chapter 9 Polymorphism and Inheritance

38 // create and return line chart.

39 }
40 }

Listing 9.14 Pseudocode for classes presenting data as bar chart and line chart, without dupli-
cated code.

Use Inheritance to Provide Default Implementation

Inheritance can be used to specify a default implementation of a method in an interface. Con-

sider for example the interface java.awt.event.MouseListener, which shall be imple-
mented by a class that is to be notified when the user does something with the mouse. This
interface has five methods, for pressing a mouse button, releasing a button, clicking a button,
moving the cursor into a window on the screen and moving it out of the window. A class
that is interested in mouse clicks would implement MouseListener and provide an imple-
mentation of the method mouseClicked. However, since the interface is implemented, that
class would also have to provide empty implementations of the other four methods. These
four empty methods will make the class more difficult to understand, and also lower its cohe-
sion. To remedy this problem, there is a class called MouseAdapter, which provides empty
implementations of all the five methods. Our class that reacts to mouse clicks can then extend
MouseAdapter, instead of implementing MouseListener, thus not having to include the
four empty methods.
Note that it is not a requirement for a default method implementation to be empty, such
a method could instead provide a frequently used implementation. Since Java 8, an inter-
face itself can have a default method implementation. This might suggest that a class like
MouseAdapter is not needed, the implementations can instead be placed in the interface it-
self. Even though this is true, it is generally not a good idea. The purpose of an interface
is to provide a contract in the form of a method declaration, nothing more. There is a clear
separation of public interface and implementation if the interface contains only declarations,
but not if it also contains implementations. In fact, the possibility to provide a default imple-
mentation in an interface was never intended to be used like this. As is stated in the Oracle
documentation[Java Tutorial], the purpose of default implementation is to make it possible to
add methods to an existing interface, without breaking the code of all classes that implement
a previous version of that interface.

When Class Hierarchies Are Appropriate

In spite of the critique of classical inheritance hierarchies above, they can still be useful. The
point is they should not be used blindly, but only under the right circumstances. First of all the
following conditions must be met.

1. All members of the superclass must be meaningful also in the subclass.

2. The superclass is a generalization of the subclass, it is more abstract than the
subclass, and can therefore be used in more situations than the subclass.

190
 Chapter 9 Polymorphism and Inheritance

3. The subclass is a specialization of the superclass. It can not be used in all

situations where the superclass can be used, but where it can be used, it is
better suited than the superclass.
4. The subclass and the superclass must have an is-a relation, meaning the ex-
pression a <subclass name> is a <superclass name> must be true.

Unfortunately, the above conditions are only necessary, but not sufficient. In fact, the hi-
erarchy of animals above, which prohibited code reuse, does meet all four requirements. To
formulate a sufficient condition is notably more difficult. A rough rule of thumb could be
that inheritance hierarchies are not very useful for real world entities, such as the animals.
The reason is that such entities are complex, and do not always follow a strict generalization-
specialization tree structure. Instead, inheritance is more useful for purely fabricated entities,
that do not really exist in the real world.
An example of a well-functioning hierarchy, not modeling real world entities, is the collec-
tion api in java.util. A small subset of the classes in this api is illustrated in figure 9.10.
For example, there is the interface List, which defines the contract of a list. This interface
is implemented by AbstractList, which provides code common for both linked lists and
array lists. Then, the classes ArrayList and LinkedList contain code that is specific for a
particular kind of list. When creating these classes, the author is free to decide which classes
shall exist, how they relate to each other, and which functionality to put where. With the ani-
mals, there is no such freedom. A penguin is a bird and a salmon is a fish, there is no arguing
about that.

Figure 9.10 Part of the collection api in java.util, which is an example of a class hierarchy modeling
fabricated entities instead of real world entities.

191
 Chapter 9 Polymorphism and Inheritance

9.4 Gang of Four Design Patterns

A pattern is a common and proven solution to a reoccurring problem. Typically, developers
realize that a particular problem in software development is solved many times, in different
programs, but the solution is always more or less the same. If this solution works well, it is
worth creating a formalized description covering the problem, variants of the solution, advan-
tages and disadvantages of the solution, etc. This formalized description is a pattern. If, as is
the case here, the solution concerns design, it is a design pattern. A collection of patterns is
like a cookbook for software development.
Gang of Four, or GoF, refers to the book Design Patterns: Elements of Reusable Object-
Oriented Software by Gamma, Helm, Johnson and Vlissides [GOF]. This book contains a
collection of patterns, often called GoF patterns. These patterns are very common, and all
developers should be comfortable using at least the more common ones. Knowledge of these
patterns also forms a vocabulary for software developers, that can be used to discuss solutions
to different problems. Thus, even though they are more advanced and abstract than what we
have seen so far, and may at first seem a bit scary, it is fundamental to master at least a few of
the 23 patterns covered in the book. This section provides a short introduction, it covers five
patterns, plus a simplified version of a sixth.

The Observer Pattern

The observer pattern is used where objects of one
or more classes must be informed whenever a par-
ticular object changes state. This is one of the
most frequently used GoF patterns, all kinds of
listeners and event handlers are variants of the ob-
server pattern.
Figure 9.11 Using the observer pattern, the ob-
serving objects are immediately notified when
Overview the observed object changes state. Image by un-
known creator [Public domain], via https://pixabay.com

The idea is to hide the observer object from

the observed class with an interface, figure 9.12.
Now, the observed class can call the observer
when it changes state, but it does not know the
class of the observer, it only knows the name of
the interface. In fact, the observed class in list-
ing 9.15 does not contain the name of the class
that implements the Observer interface. Using
an interface this way is a good practice when the
caller is interested only in a particular ability of
the callee. In this case, the observed class needs
Figure 9.12 General class diagram for any ob- a class with the ability to receive notifications
server pattern implementation. The observed about updates, which is provided by the notify
class has no reference to the observing class,
but only to the Observer interface.

192
 Chapter 9 Polymorphism and Inheritance

method. The observed class does not care what else the observing object can do, or what its
purpose is, as long as it can receive notifications.

Case Study

Here, an observer will be used to solve a problem we have been facing since we started to
design, namely how to pass data from model to view, when that data can not be a return value
to a method call from the view. As a case study, the car rental program is augmented with a
display on the wall in the office, telling how many cars of each size the company has rented
out. Whenever a rental is payed, no matter where or by who, the display shall be updated with
the rented car.

1 package se.leiflindback.oodbook.despat.observer;
2
3 import java.util.ArrayList;
4 import java.util.List;
5
6 /**
7 * The observed class in a general implementation of the
8 * observer pattern.
9 */
10 public class ObservedClass {
11 private List<Observer> observers = new ArrayList<>();
12
13 /**
14 * Registers observers. Any <code>Observer</code> that is
15 * passed to this method will be notified when this object
16 * changes state.
17 *
18 * @param observer The observer that shall be registered.
19 */
20 public void addObserver(Observer observer) {
21 observers.add(observer);
22 }
23
24 // Called by any method in this class that has changed
25 // the class’ state.
26 private void notifyObservers() {
27 for (Observer observer : observers) {
28 observer.stateHasChanged();
29 }
30 }
31 }

Listing 9.15 The observed class, note that there is no reference to the observing class, only to the
Observer interface.

193
 Chapter 9 Polymorphism and Inheritance

There are many reasons why this is difficult to implement. It is not easy to handle the
update directly in the view, letting the user interface where the rental is payed update the
display showing rented out cars. First, there might be many different ways to rent a car, in
different locations, and they can not all know about the display. Second, even if there is only
one single place where a customer can rent a car, it is still bad design to let one object in the
view be responsible for updating other view objects. That knowledge, what to do in response
to a particular user action, is business logic, and should therefore not be handled by the view
itself. A view object shall only call the controller and update itself to reflect the result of
the call. Finally, in the current rent car implementation, the view does not keep track of the
rental being paid for, that is done by the controller. There is no reason to give the view the
responsibility to know the size of the rented car, that would reduce cohesion in the view, since
it is the responsibility of Rental, in the model. Also, there would certainly be some amount
of duplicated code if both Rental and view had to know which car was being rented.
Having established that the display can not be updated by view object(s) responsible for the
user interface where the car is rented, we face a brick wall. There is no other object in the
view, and it would ruin MVC if the controller or any lower layer called the display to update
the number of rented cars. The solution is to use an observer, which will allow the model to
call the view without any dependency on the view.

Solution

First, we create the observer interface. It is convenient to include observer in the

name, to clearly state it is part of the observer pattern. It could be called for example
CarIsHiredObserver, but let’s be a bit less specific and call it RentalObserver. That
way it can, in the future, handle information also about other state changes. Being too specific
always increases the risk of change. The notification method can be called newRental, since
it tells that a new rental is completed. The interface is listed in listing 9.16.

1 package se.leiflindback.oodbook.rentcarWithExAndDesPat.model;
2
3 /**
4 * A listener interface for receiving notifications about
5 * rented cars. The class that is interested in such
6 * notifications implements this interface, and the object
7 * created with that class is registered with
8 * <code>addRentalObserver</code>. When a car is rented, that
9 * object’s <code>newRental</code> method is invoked.
10 */
11 public interface RentalObserver {
12 /**
13 * Invoked when a rental has been paid.
14 *
15 * @param rentedCar The car that was rented.
16 */
17 void newRental(CarDTO rentedCar);

194
 Chapter 9 Polymorphism and Inheritance

18 }

Listing 9.16 The observer interface, RentalObserver.

The next decision is which class to observe. The required notification is to tell which
type of car was rented when a payment is accepted. Therefore, the observed class must have
knowledge about both payment and rented car. These requirements are met by Rental, which
is also chosen. To make it observable, it must have a method that registers observers. This
method is called addObserver, just as in the general example in listing 9.15. Also, it is good
practice to add a private method, notifyObservers, which is called by any other method
when notifications shall be sent. When shall observers be notified? In this case, the only state
change of interest is that a payment is accepted, therefore, notifyObservers is called last
in the method pay. Listing 9.17 contains the updated parts of Rental.

1 package se.leiflindback.oodbook.rentcarWithExAndDesPat.model;
2
3 public class Rental {
4 private List<RentalObserver> rentalObservers =
5 new ArrayList<>();
6
7 public void pay(CashPayment payment) {
8 payment.calculateTotalCost(this);
9 this.payment = payment;
10 notifyObservers();
11 }
12
13 private void notifyObservers() {
14 for (RentalObserver obs : rentalObservers) {
15 obs.newRental(rentedCar);
16 }
17 }
18
19 public void addRentalObserver(RentalObserver obs) {
20 rentalObservers.add(obs);
21 }
22 }

Listing 9.17 The observed class, Rental.

Then it is time to create the observer implementation. It shall be a class in the view, sim-
ulating the display telling the number of rented cars. Currently, there is only one class in the
view, which contains the hardcoded sample execution. To improve cohesion, a new class,
RentedCarsDisplay, is created. This class implements RentalObserver, and simply
prints the required information to System.out, see listing 9.18. As can be seen on line 28, it
was appropriate to include a CarDTO representing the rented car as parameter in the notifica-
tion method, newRental, since CarDTO contains the required information.

195
 Chapter 9 Polymorphism and Inheritance

Previously, the size of the rented car was represented as a String. That was never really
appropriate, due to the risk of misspelling. Now that the size is being used on many places in
the code, this problem is growing. Therefore, size is changed to be defined using an enumera-
tion. This enumeration is placed in CarDTO, see listing 9.19, to make it available everywhere
a CarDTO is used.

1 package se.leiflindback.oodbook.rentcarWithExAndDesPat.view;
2
3 /**
4 * Shows a running total of rented cars of each type.
5 */
6 class RentedCarsDisplay implements RentalObserver {
7 private Map<CarDTO.CarType, Integer> noOfRentedCars =
8 new HashMap<>();
9
10 /**
11 * Creates a new instance, with the all counters of rented
12 * cars set to zero.
13 */
14 public RentedCarsDisplay() {
15 for (CarDTO.CarType type : CarDTO.CarType.values()) {
16 noOfRentedCars.put(type, 0);
17 }
18 }
19
20 @Override
21 public void newRental(CarDTO rentedCar) {
22 addNewRental(rentedCar);
23 printCurrentState();
24 }
25
26 private void addNewRental(CarDTO rentedCar) {
27 int noOfRentedCarsOfThisType =
28 noOfRentedCars.get(rentedCar.getSize()) + 1;
29 noOfRentedCars.put(rentedCar.getSize(),
30 noOfRentedCarsOfThisType);
31 }
32
33 private void printCurrentState() {
34 System.out.println("### We have now rented out ###");
35 for (CarDTO.CarType type : CarDTO.CarType.values()) {
36 System.out.print(noOfRentedCars.get(type));
37 System.out.print(" ");
38 System.out.print(type.toString().toLowerCase());
39 System.out.println(" cars.");
40 }

196
 Chapter 9 Polymorphism and Inheritance

41 System.out.println("##############################");
42 }
43 }

Listing 9.18 The observer implementation, RentedCarsDisplay.

1 public final class CarDTO {

2 // Unchanged code
3 public enum CarType {SMALL, MEDIUM, LARGE};
4 // Unchanged code
5 }

Listing 9.19 The definition of the enum in CarDTO.

The last part of the observer implementation is to create an instance of RentedCarsDisplay,

and register it with Rental. There are really only two candidates for creating this object, the
startup class Main and the view placeholder View. Creating it in Main creates more depen-
dencies between the layers startup and view, which means unnecessary coupling. It is
better to let Main create only one object in the view, as is the case now, and then let that object
unfold the entire view. This unfolding means to create the RentedCarsDisplay object.
How shall the view pass the observer to the observed class in the model? There is only
one allowed way for a view to communicate with a model, via controller. The solution is thus
to pass the observer from view to controller, and then from controller on to Rental in the
model, see listings 9.20 and 9.21. There are two things to note in the last listing. First there is
no need to register the observer again for each new rental. It is instead saved in Controller
and registered to all future Rentals. Second, since there might be more than one observer, the
controller keeps all of them stored in a list, and registers them all when a new rental is initiated.
That means Rental needs a new method, addObservers, which allows registering a whole
list of observers.

1 public class View {

2 // Unchanged code
3 public View(Controller contr) throws IOException {
4 this.contr = contr;
5 contr.addRentalObserver(new RentedCarsDisplay());
6 this.logger = new LogHandler();
7 }
8 // Unchanged code
9 }

Listing 9.20 Creating the observer in View.

1 public class Controller {

2 private List<RentalObserver> rentalObservers =

197
 Chapter 9 Polymorphism and Inheritance

3 new ArrayList<>();
4
5 public void registerCustomer(CustomerDTO customer) {
6 // Unchanged code.
7 rental.addRentalObservers(rentalObservers);
8 }
9
10 /**
11 * The specified observer will be notified when a rental
12 * has been paid. There will be notifications only for
13 * rentals that are started after this method is called.
14 *
15 * @param obs The observer to notify.
16 */
17 public void addRentalObserver(RentalObserver obs) {
18 rentalObservers.add(obs);
19 }
20 }

Listing 9.21 Handling observer registration in Controller.

Comments

There is much worth pondering about the observer pattern. Here are some issues concerning
the implementation.

• What shall the observer know about the observed class? The observer
knows nothing at all about the observed class in the case study above. In fact,
there is no way for it to get a reference to the object that sent the notification.
Instead, all relevant data is passed as a parameter in the call of the notification
method, in this case, the data is a CarDTO object. This solution becomes
more and more problematic as more data is required. Also, whenever the
need for data changes, the parameter list of the notification method must
change, which means the public interface of Observer is changed.
Another solution is to not pass any data at all, but instead a reference to the
object that changed state, the Rental object in the above case study. The ob-
server can then call get methods in the observed class to collect needed data.
This solution is more dynamic, since different observers can collect different
data, as required. Also, the parameter list of the notification method consists
of one single parameter, the object that changed state, and will never change.
The downside of such an implementation is that the observer now depends
on the observed class, thereby increasing coupling. To alleviate this prob-
lem, we can introduce yet one interface, call it Observed, implemented by
the observed class. This interface shall contain only methods of the observed

198
 Chapter 9 Polymorphism and Inheritance

class allowed to call by the observer. Now, the observer will know only about
this interface, not about the actual class being observed.

• Is an event class needed? It is sometimes appropriate to introduce a new

class, representing the actual event. If so, an object of this class is normally
the only parameter of the notification method. Any data can be included in
this object, and the parameter list of the notification method is thus no longer
cluttered with this data. The event object can also contain a reference to the
observed object, if that is needed, and can it contain any other information
of interest concerning the event.

• When shall observers be notified? Observers must be informed about a

particular event only when that event is really completed. It would for ex-
ample not be acceptable to call notifyObservers on line eight, instead of
line ten, in listing 9.17, since the payment has not yet been handled on line
eight.

• Broadcasted method call! Using an observer does, in fact, enable broad-

casting a method call. Instead of calling a method in one observer object,
the observed object calls notifyObservers, which means the method is
called in many observer objects. This really powerful, but, as many powerful
solutions, carry a risk. The risk is that the caller can not know how many
objects are called, nor what all those objects will do when called. Therefore,
what seems like an innocent method call to the observed class, might in real-
ity result in lengthy and resource consuming operations. To avoid this, event
handling methods in observers shall not perform such costly operations.

The Strategy Pattern

The strategy pattern makes it possible to swap
an algorithm without changing existing code. In
fact, the logging API created in section 9.2 used
strategy to change logger implementation. It was
possible to change between FileLogger and
ConsoleLogger while the program was run-
ning, and it was also possible to add new classes
implementing Logger without any changes to the
code. Algorithm is used here in a very wide sense,
it can be any kind of behavior.
Figure 9.13 Using the strategy pattern, any algo-
rithm can be plugged in to an existing pro-
gram. Image by unknown creator [Public domain], via
http://www.publicdomainpictures.net

199
 Chapter 9 Polymorphism and Inheritance

Overview

The strategy pattern is a perfect example of

the benefits of polymorphism. If the client
using the algorithm depends only on the in-
terface, and not on the implementations, as
in figure 9.14, it is possible to change imple-
mentation without any changes to the client.
Figure 9.13 also illustrates this, both the
Figure 9.14 The client can use any class that imple- red and the blue piece fit. The surrounding
ments StrategyDefinition, it is completely
independent of the implementing classes.
pieces are not affected as long as the shape
(definition) is correct.

Case Study

The logging API from section 9.2 would be a great case study. Still it is not used, since there
are already many existing logging APIs. It would be a bit inappropriate to spend a lot of effort
developing yet another. The case study used here is instead the search for an available car
matching the wishes of a customer. This is performed in CarRegistry, where the current
search algorithm, listing 9.22, considers a car to be a match if all properties except registration
number are equal to those of the searched car. Properties equal to null or zero are ignored,
and the matching car must not be booked. This is, however, only one of many possible ways
to match existing cars against a customer’s wishes.

1 /**
2 * Search for a car that is not booked, and that matches the
3 * specified search criteria.
4 *
5 * @param searchedCar This object contains the search
6 * criteria. Fields in the object that are
7 * set to <code>null</code> or zero are
8 * ignored.
9 * @return A description matching the searched car’s
10 * description if an unbooked car with the
11 * same features as <code>searchedCar</code> was
12 * found, <code>null</code> if no such car was found.
13 * /
14 public CarDTO findAvailableCar(CarDTO searchedCar) {
15 for (CarData car : cars) {
16 if (matches(car, searchedCar)) {
17 return new CarDTO(car.regNo,
18 new Amount(car.price), car.size,
19 car.AC, car.fourWD, car.color,
20 false);
21 }
22 }

200
 Chapter 9 Polymorphism and Inheritance

23 return null;
24 }
25
26 private boolean matches(CarData found, CarDTO searched) {
27 if (searched.getPrice() != null &&
28 !searched.getPrice().equals(new Amount(found.price))){
29 return false;
30 }
31 if (searched.getSize() != null &&
32 !searched.getSize().equals(found.size)) {
33 return false;
34 }
35 if (searched.getColor() != null &&
36 !searched.getColor().equals(found.color)) {
37 return false;
38 }
39 if (searched.isAC() != found.AC) {
40 return false;
41 }
42 if (searched.isFourWD() != found.fourWD) {
43 return false;
44 }
45 if (found.booked) {
46 return false;
47 }
48 return true;
49 }

Listing 9.22 The algorithm used to match cars, before using the strategy pattern.

Now, it shall be possible to switch between multiple, different, matching algorithms. As a

case study, we keep the existing algorithm, and also add two new. One exact match, where
all properties of a car in the registry must be equal to those of the searched car, and one
algorithm where a specific available car becomes the match, if only one of its properties
matches the search criteria. The latter algorithm is used when the rental company wants to
promote a particular car. It would be a very bad solution to hardcode the different algorithms
in CarRegistry. With that approach, the code would be very messy, and it would be difficult
to change algorithm. Instead, a strategy interface will be created, and CarRegistry will be
assigned an implementation of that interface performing the desired search algorithm.

Solution

Figure 9.15 shows how a strategy interface, Matcher, and its implementations are used to
handle the search for a car in the car registry. Note that CarRegistry has no dependency on
the implementations, only on the interface. As can be seen from the definition of the match
method, it is handed a list containing all available cars. This approach can be questioned, is it

201
 Chapter 9 Polymorphism and Inheritance

not time consuming to read all cars from the database and place them in a list? Maybe yes, the
search could have been defined in the database query, in a way that only a matching car was
returned. To solve the problem that way, match would have needed a database connection for
sending the appropriate query to the database, instead of the list of all existing cars. There are,
however, reasons to solve the problem as in figure 9.15 instead. Above all, there is no database
yet. Also, even though the proposed solution is more time consuming than a database query, it
will probably not create problems, since the list can not be extremely long. It is unlikely that
the rental company has more than a few hundred available cars to choose from. Another thing
worth noting is that an available car is specified by a CarDTO object, not by the CarData
object used internally in CarRegistry. This is to maintain the distinction that CarData
is a a replacement for a database, it is not appropriate for use outside CarRegistry. The
Matcher interface is listed in listing 9.23.

Figure 9.15 The strategy case study. There are three different algorithms for matching a car in the registry against
a customer’s wishes.

1 /**
2 * Defines the ability to match existing cars with a searched
3 * car. This interface shall be implemented by a class that
4 * provides a matching algorithm.
5 */
6 public interface Matcher {
7 /**
8 * Searches the specified available cars for an instance
9 * matching the specified search criteria.
10 *
11 * @param searched Search criteria
12 * @param available Available cars
13 * @return A matching car, or <code>null</code> if none
14 * was found.
15 */
16 CarDTO match(CarDTO searched, List<CarDTO> available);
17 }

Listing 9.23 The Matcher interface, wich defines the matching algorithm.

202
 Chapter 9 Polymorphism and Inheritance

Next, it is time to create the implementations. The existing search algorithm looks for a
car with all properties, except registration number, equal to the properties of the searched car.
However, searched properties equal to null or zero are ignored. Therefore, this algorithm is
called WildCardMatch. The algorithm that never ignores any property, but requires all prop-
erties except registration number to match, is called PerfectMatch. Finally, the algorithm
used to promote a particular car is called PromotingMatch. The first two can be found in the
accompanying GitHub repository [Code], only the last is listed here, listing 9.24. The registra-
tion number of the car to promote is specified in setCarToPromote, lines 21-23. The search
algorithm is, of course, written in the method match, defined by the implemented interface.
PerfectMatch is used as a fallback if the promoted car is different from the searched car in
all aspects (line 59). This is in order to not have to tell the customer “sorry, there is no car”,
even if there actually is one.

1 /**
2 * A <code>Matcher</code> that finds the car that shall be
3 * promoted, provided it has at least one property, except
4 * registration number, matching the search criteria. If it
5 * has not, performs a <code>PerfectMatch</code>.
6 */
7 public class PromotingMatch implements Matcher {
8 private String regNoOfCarToPromote;
9
10 PromotingMatch() {
11 }
12
13 /**
14 * Specify which car to promote.
15 *
16 * @param regNo The car with this registration number will
17 * be found by the matching algorithm, if it
18 * exists and has at least one property equal
19 * to the search criteria.
20 */
21 public void setCarToPromote(String regNo) {
22 this.regNoOfCarToPromote = regNo;
23 }
24
25 @Override
26 public CarDTO match(CarDTO searched,
27 List<CarDTO> available) {
28 for (CarDTO carToMatch : available) {
29 if (!regNoOfCarToPromote.
30 equals(carToMatch.getRegNo())) {
31 continue;
32 }
33 if (carToMatch.getPrice() != null &&

203
 Chapter 9 Polymorphism and Inheritance

34 searched.getPrice() != null &&

35 searched.getPrice().equals(
36 carToMatch.getPrice())) {
37 return carToMatch;
38 }
39 if (carToMatch.getSize() != null &&
40 searched.getSize() != null &&
41 searched.getSize().equals(
42 carToMatch.getSize())) {
43 return carToMatch;
44 }
45 if (carToMatch.getColor() != null &&
46 searched.getColor() != null &&
47 searched.getColor().equals(
48 carToMatch.getColor())) {
49 return carToMatch;
50 }
51 if (searched.isAC() == carToMatch.isAC()) {
52 return carToMatch;
53 }
54 if (searched.isFourWD() ==
55 carToMatch.isFourWD()) {
56 return carToMatch;
57 }
58 }
59 return new PerfectMatch().match(searched, available);
60 }
61 }
Listing 9.24 PromotingMatch, wich performs a matching algorithm that tries to match a partic-
ular car.

The last part of the strategy pattern is the client, CarRegistry. The use of the matching
algorithm can be found on line 25 in listing 9.25. Unfortunately, there is a dependency on the
implementation of the algorithm, not only on the definition, since the WildCardMatch object
is created on this line. It is still worth the effort to define the Matcher interface, since the
only change required to switch algorithm is to change WildCardMatch for the desired class
name. However, it is not at all as beautiful as the implementation of the logging algorithm in
section 9.2, where it was possible to change algorithm while the program was running. This
defect will be mitigated with the next pattern, factory.

1 /**
2 * Search for a car that is not booked, and that matches the
3 * specified search criteria.
4 *
5 * @param searchedCar This object contains the search
6 * criteria. Fields in the object that are

204
 Chapter 9 Polymorphism and Inheritance

7 * set to <code>null</code> or zero are

8 * ignored.
9 * @return A description matching the searched car’s
10 * description if an unbooked car with the
11 * same features as <code>searchedCar</code> was
12 * found, <code>null</code> if no such car was found.
13 */
14 public CarDTO findAvailableCar(CarDTO searchedCar) {
15 List<CarDTO> allCars = new ArrayList<>();
16 for (CarData car : cars) {
17 if (!car.booked) {
18 allCars.add(new CarDTO(car.regNo,
19 new Amount(car.price),
20 car.size, car.AC,
21 car.fourWD, car.color,
22 car.booked));
23 }
24 }
25 return new WildCardMatch().match(searchedCar, allCars);
26 }

Listing 9.25 The call from CarRegistry to the matching algorithm.

Comments

• Algorithms might need a state Concrete strategies might differ only in

state, that is, the value of some property. If so, there is no need for a new
class for each state. An example of this is PromotingMatch, which tries
to match a particular car. Which car to promote is specified as a registration
number. There is no need to create a new class just to promote another car.

• All algorithms have the same public interface It is problematic that all con-
crete strategies implement the same interface, if different implementations of
the algorithm need different data. An example in the case study could be if
some matching algorithms need to call the database, while others do not. In
this case, a database connection must be passed to the match method, even
though some classes will never use it. This might seem a small problem, the
solution is to simply set the database connection parameter to null, when
using an algorithm that does not call the database. However, it is not high
cohesion, nor a very nice public interface, to have methods with parameters
they do not use. The bigger the public interface, and the more the algorithms
differ, the bigger this problem gets. Unfortunately, there is no solution except
to unify the algorithm’s public interfaces as much as possible.

• Who shall instantiate implementations? As stated above, it is unfortunate

205
 Chapter 9 Polymorphism and Inheritance

that, in the case study, there is a dependency on the implementation of the

algorithm, not only on the definition. This dependency can be found on line
25 in listing 9.25, where the concrete strategy is instantiated. This instanti-
ation problem is relatively common when using the strategy pattern, and a
common solution is to use a factory, which is the next pattern.

The Factory Pattern

In some cases, instantiating objects worsens the
design, wherever it is done. The reason could
be that an unwanted coupling is created, as is
the case above, where CarRegistry depends
on the strategy implementations. It could also
be that creating an object is complicated, and
thereby gives bad cohesion to the class where it
is placed. Whatever the reason why instantiat-
ing objects worsens the design, the solution is to
create a completely new class, whose sole pur-
pose is to create those objects. This new class is
called a factory. There are, in fact, many GoF pat- Figure 9.16 The factory delivers the desired prod-
terns concerning instantiation, but none of them uct. They all look the same on the outside (pub-
is called Factory. The pattern described here is a lic interface), but the factory knows which has
simplified version of some of those GoF patterns. the appropriate content (implementation).

Overview

As is illustrated in figures 9.16 and 9.17, the

client does not know the class of the object
it gets from the factory. The client simply
asks for a product, and happily uses what-
ever concrete product it is handed. All that
matters is that this concrete product can per-
form the work specified by the Product in-
terface. The question that immediately arises
is “How can the factory know which concrete
product is appropriate for the client?” Unfor-
tunately, there is no simple answer, but some
Figure 9.17 Client has no dependency on the concrete possibilities are discussed below.
products, when they are created by the Factory.

Case Study

Now it is finally time to remove CarRegistry’s dependency on the concrete matching algo-
rithms, which has been troubling us since it was introduced on line 25 in listing 9.25. It is
not a good design that CarRegistry must be changed and recompiled, when the algorithm

206
 Chapter 9 Polymorphism and Inheritance

matching a customers wishes against available cars is swapped. The whole purpose of using
the strategy pattern was to make CarRegistry independent of those algorithms.

Solution

The solution to the undesired dependency problem is to introduce a factory. Let’s

call it MatcherFactory, since it creates matching algorithms. It needs only one
method, which shall return an object of a class implementing Matcher, defined in list-
ing 9.23. This class shall implement the method match, which performs the actual
matching algorithm. The open question is how the factory shall know which of the
concrete matchers to return. The solution used here, is to let the factory read the
class name of the Matcher implementation from a system property, see line 31-32 in
listing 9.26. A system property is specified on the command line with the switch
-D. If, for example, the car rental program is started with a command line containing
-Dse.leiflindback.rentcar.matcher.classname=se.leiflindback.oodbook.
rentcarWithExAndDesPat.integration.matching.WildCardMatch, the value re-
turned by System.getProperty on line 32 would be se.leiflindback.oodbook.
rentcarWithExAndDesPat.integration.matching.WildCardMatch, which is the
class name of the desired implementation.
System properties constitute a nice way to pass initialization data to the object need-
ing it. A system property can be specified on the command line, and read using
System.getProperty anywhere in the code. It is not necessary to pass a system property
from main to the class using it, as is the case with a command line parameter.

1 /**
2 * A Factory that creates instances of the algorithm used
3 * for matching a customer’s wishes against available cars.
4 * All such instances must implement <code>Matcher</code>.
5 */
6 public class MatcherFactory {
7 private static final String MATCHER_CLASS_NAME_KEY =
8 "se.leiflindback.rentcar.matcher.classname";
9 private static final String CAR_TO_PROMOTE_KEY =
10 "se.leiflindback.rentcar.matcher.promote";
11
12 /**
13 * Returns a <code>Matcher</code> performing the default
14 * matching algorithm. The class name of the default
15 * <code>Matcher</code> implementation is read from the
16 * system property
17 * <code>se.leiflindback.rentcar.matcher.classname</code>
18 *
19 * @return The default matcher
20 * @throws ClassNotFoundException If unable to load the
21 * default matcher class.

207
 Chapter 9 Polymorphism and Inheritance

22 * @throws InstantiationException If unable to instantiate

23 * the default matcher class.
24 * @throws IllegalAccessException If unable to instantiate
25 * the default matcher class.
26 */
27 public Matcher getDefaultMatcher() throws
28 ClassNotFoundException,
29 InstantiationException,
30 IllegalAccessException {
31 String className =
32 System.getProperty(MATCHER_CLASS_NAME_KEY);
33 Matcher matcher =
34 (Matcher) Class.forName(className).newInstance();
35 if (matcher instanceof PromotingMatch) {
36 ((PromotingMatch)matcher).
37 setCarToPromote(System.
38 getProperty(CAR_TO_PROMOTE_KEY));
39 }
40 return matcher;
41 }
42 }

Listing 9.26 The MatcherFactory class, which hands out a matching algorithm.

After the class name of the appropriate matcher is read from the system property, it still
remains to create an object of that class. Remember, from listing 9.8, that it is not required to
know the name of the class to be instantiated at compile time. It is possible to create an object
of a class, whose name is the content of a String variable. This is done on lines 33-34 in
listing 9.26.
The goal of removing all references to Matcher implementations from CarRegistry has
been reached now that the factory has been introduced, as can be seen from listing 9.27. This
has taken us closer to the point where we can change matching algorithm without having
to restart the program, but not all the way there. To completely reach that goal, we need a
mechanism to first change the value of the system property, and then to force the factory to
read the new value. How to do this is discussed in the Comments subsection, below.

1 /**
2 * Search for a car that is not booked, and that matches the
3 * specified search criteria.
4 *
5 * @param searchedCar This object contains the search
6 * criteria. Fields in the object that are
7 * set to <code>null</code> or zero are
8 * ignored.
9 * @return A description matching the searched car’s
10 * description if an unbooked car with the

208
 Chapter 9 Polymorphism and Inheritance

11 * same features as <code>searchedCar</code> was

12 * found, <code>null</code> if no such car was found.
13 * @throws CarRegistryException If unable to search for a
14 * matching car.
15 */
16 public CarDTO findAvailableCar(CarDTO searchedCar) {
17 List<CarDTO> allCars = new ArrayList<>();
18 for (CarData car : cars) {
19 if (!car.booked) {
20 allCars.add(new CarDTO(car.regNo,
21 new Amount(car.price), car.size,
22 car.AC, car.fourWD, car.color,
23 car.booked));
24 }
25 }
26 try {
27 return new MatcherFactory().getDefaultMatcher().
28 match(searchedCar, allCars);
29 } catch (ClassNotFoundException | InstantiationException |
30 IllegalAccessException ex) {
31 throw new CarRegistryException(
32 "Unable to instantiate matcher", ex);
33 }
34 }

Listing 9.27 The call from CarRegistry to the factory (line 27). There is no dependency on the
Matcher implementations

Comments

• How shall the factory know what to create? There are many different ways
to configure the factory, for example using a system property, as in the case
study above. Other alternatives are to have the factory read from a file, or to
create a user interface where the product can be specified. It might also be ap-
propriate to add a method to the factory, that can be used to define the product
to create, for example public void setDefaultProduct(Product
product).
No matter how the class name of the desired product is communicated to
the factory, there must also be a way to tell the factory to read it. It is not
enough to set product when the program is started, we also want to be able
to change during execution. If this change is to be initiated by something
outside the program, input must be accepted. There could for example be a
user interface which allows a system administrator to make the factory read
a new value. If the product’s class name is stored in the factory, it must have
a method that can be called to force it to re-read that class name.

209
 Chapter 9 Polymorphism and Inheritance

The product created by the factory in the case study did not depend on pro-
gram state in any way. All matching algorithms were always allowed, no
matter which cars were available or what the customer wished. This is
not always the case, we could for example add another factory, that cre-
ates products for payment handling. If the customer pays cash, the fac-
tory shall create a CashPayment, if the customer pays with credit card, a
CreditCardPayment shall be created. In this case, there is only one al-
lowed product. That means the factory’s method that creates a product
must take a parameter that enables deciding which is the correct product.
The method definition could be for example public PaymentHandler
createPaymentHandler(PaymentType type).

• Try to make the factory independent of concrete products. One of the

concrete products in the case study requires extra initialization data, namely
PromotingMatch, which needs the registration number of the car to pro-
mote. This data is passed to the instance on lines 36-38 in listing 9.26 above.
There is a problem with these lines, the class name PromotingMatch
is mentioned, creating a coupling to a concrete product. That is a dan-
gerous road to take, dependency on concrete products means the code
must be changed when new products are added, or existing products are
changed. We could get rid of this dependency by passing initialization
data to all concrete products, whether they need it or not. A new method,
void init(Map<String, Object> initData), should be added to
the Matcher interface. This method should have a java.util.Map pa-
rameter, with key-value pairs containing all data needed by the matching al-
gorithm. PromotingMatch then requires such a map where the registration
number is specified, for example with the key REG_NO. The other matching
algorithms, PerfectMatch and WildCardMatch, do not need init parame-
ters and would therefore have empty init methods. If we made this change,
lines 33-40 of listing 9.26 would instead be as in listing 9.28, where there is
no dependency on any concrete product.

1 Matcher matcher =
2 (Matcher)Class.forName(className).newInstance();
3 //read initialization data, for example from
4 //a file, and place it in a java.util.Map
5 //called initData.
6 matcher.init(initData);
7 return matcher;

Listing 9.28 Configuration of products without dependency on any concrete

product

• Products can be cached. The suggested solution above creates a new con-
crete product each time getDefaultMatcher is called. If the product is

210
 Chapter 9 Polymorphism and Inheritance

needed often and is time consuming to create, it might be a better idea to

save one instance of each class. Those instances could be created when the
factory is created, in order not having to spend any time on creation when a
product is requested.

• Who creates the factory itself? In listing 9.27, line 27, CarRegistry
creates a new instance of the factory each time it is needed. That is fine in
the case study program, but what if it is time consuming to create the factory,
maybe because it creates cached instances of concrete products as suggested
in the previous bullet? And what if the factory reads a configuration file as
suggested above, which is also time consuming? If, for any reason, it is not
appropriate to create new factory instances, there must be a way to guarantee
that there is only one, single, instance. There must also be a way to share
this instance between all methods where it is needed. This is the purpose of
the next pattern.

The Singleton Pattern

Sometimes it is not possible to allow more than
one instance of a particular class. An example is
the logging API created in section 9.2. If there
is more than one instance of ConsoleLogger,
the outputs of the loggers will be mixed on the
screen, and if there is more than one instance of
FileLogger, the outputs will be mixed in the
file. To handle this situation, the program must
both inhibit creating more objects, and enable all
classes needing the single instance to get a refer-
ence to it. Figure 9.18 A singleton set has only one member,
there are never any more. Image by unknown creator
[Public domain], via https://pixabay.com
Overview

To block instantiation, the constructor of the critical object is made private. That means it can
only be called by the class itself, no other class can create instances. Next, one single instance
is created by the class itself, and saved in a static field. Last, a method is written, which hands
out this sole instance. A class with these properties is called a singleton. The class diagram in
figure 9.19a illustrates these features. Remember that underlining a member means it is static.
Since any singleton, by definition, has such a static method and field, and private constructor,
they are sometimes omitted and instead indicated with a stereotype, as in figure 9.19b.

Case Study

As a case study of the singleton pattern, consider the factory class created above. A factory is
often a candidate to become a singleton, especially if the objects it creates shall be cached.

211
 Chapter 9 Polymorphism and Inheritance

(a) (b)

Figure 9.19 Class diagram illustrating a singleton.

(a) The static method and field, and the private constructor are depicted.
(b) The singleton-specific members from figure 9.19a are replaced with a stereotype.

Solution

The solution in listing 9.29 is pretty straightforward. The static member holding the only
existing instance is on line seven, the static method handing out this instance is on lines 12-14
and the private constructor on lines 16-17. Listing 9.30 shows how the client, CarRegistry,
calls getFactory on line 12, in order to retrieve the only existing instance of the singleton.

1 /**
2 * A Singleton that creates instances of the algorithm used
3 * for matching a customer’s wishes against available cars.
4 * All such instances must implement <code>Matcher</code>.
5 */
6 public class MatcherFactory {
7 private static final MatcherFactory MATCHER_FACTORY =
8 new MatcherFactory();
9 /**
10 * @return The only instance of this singleton.
11 */
12 public static MatcherFactory getFactory() {
13 return MATCHER_FACTORY;
14 }
15
16 private MatcherFactory() {
17 }
18
19 // The getDefaultMatcher method has not been changed from
20 // listing 9.26.
21 }

Listing 9.29 The MatcherFactory class, from listing 9.26, after it has been turned into a single-
ton.

1 public CarDTO findAvailableCar(CarDTO searchedCar) {

2 List<CarDTO> allCars = new ArrayList<>();
3 for (CarData car : cars) {

212
 Chapter 9 Polymorphism and Inheritance

4 if (!car.booked) {
5 allCars.add(new CarDTO(car.regNo,
6 new Amount(car.price), car.size,
7 car.AC, car.fourWD, car.color,
8 car.booked));
9 }
10 }
11 try {
12 return MatcherFactory.getFactory().
13 getDefaultMatcher().match(searchedCar,
14 allCars);
15 } catch (ClassNotFoundException | InstantiationException |
16 IllegalAccessException ex) {
17 throw new CarRegistryException(
18 "Unable to instantiate matcher", ex);
19 }
20 }

Listing 9.30 CarRegistry retrieves the factory instance and calls getDefaultMatcher. The
only difference from listing 9.27 is line 12.

Comments

• Why not make all members static, instead of creating a singleton? This is
a good question that is often raised. In fact, there are many reasons, that may
not be obvious at first thought. Maybe the best reason is that a singleton is
likely to have state in the form of instance variables. If all methods are static,
then also all fields must be static, otherwise they can not be accessed by the
static methods. Already here we are giving up object orientation, but it easily
gets worse. The singleton probably has references to other objects, and when
methods in those objects are called, there is a risk that the make everything
static disease spreads also to those. Also, even if all members are made static,
we still need to add a private constructor to prohibit instantiating objects. In
fact, the question should probably be posed the opposite way, why make
something static in an object-oriented program, if it can be avoided? There
are also many other reasons, for example it is easier to change a singleton into
an ordinary object. Also, static methods can not be changed by inheritance,
and static fields can not be serialized.

• Is it never appropriate to create a class where all members are static?

Yes, in fact there is a situation where such a class is appropriate, namely
a class that represents a purely procedural API, which will never have any
state. An example of such a class is java.lang.Math, which contains
mathematical functions, like trigonometric and exponential functions.

213
 Chapter 9 Polymorphism and Inheritance

• It is impossible to pass parameters to the constructor of a singleton. Only

the singleton itself can call its private constructor. For example, the singleton
factory in the case study might have to read a configuration file to find out
which classes to instantiate. The location of this file can not be passed to the
constructor, and it seems as if it must be hard coded in the singleton itself.
There is, however, a way around this problem, at least to some extent. It is
possible to pass the file location as a system property, which can be used as
in the solution to the factory case study above.

• When is the singleton instance created? The case study singleton instance
is created when line seven in listing 9.29 is executed, but when is this? It
is when static fields are initialized, which is when a class file is loaded by
the JVM. Exactly when this happens varies, but the latest time a class can be
loaded is when the JVM encounters the class name in the program.

The Composite Pattern

The composite pattern makes it possible to treat
individual objects and groups of objects exactly
the same way. This is appropriate when a certain
task is to be performed, sometimes by just one ob-
ject, other times by each object in a set of objects.
The client shall be completely ignorant about the
difference. Client code is identical when the task
is performed once, by an individual implementa-
tion, and when it is performed multiple times, by
different implementations. Figure 9.20 The composite pattern removes the
difference between a container and what is
contained inside it. Image by unknown creator [Pub-
lic domain], via https://pixabay.com
Overview

As illustrated in figures 9.21 and 9.22, the

client has a reference to an interface, and
is ignorant about the implementations. As
many examples above have shown, this is
often the case, for example when using
the strategy pattern. What is new here, is
that there is one implementation of the in-
terface, CompositeTask, which does not
Figure 9.21 The composite implements the same interface contain any concrete implementation. In-
as the concrete tasks, and wraps a number of concrete stead, the composite has references to one
tasks. or more concrete task. When called by the
client, the composite in turn calls all its
concrete tasks, and each of those perform the task. It is then up to the composite to combine
the outcomes of the concrete tasks, and return the result to the client.

214
 Chapter 9 Polymorphism and Inheritance

Figure 9.22 When called, the composite does not perform the task, but instead calls the concrete tasks it contains.

Case Study

Consider the car matching algorithm provided by PromotingMatch, which is listed in list-
ing 9.24. This algorithm tries to match the searched car with the car that shall be pro-
moted, and, if it fails, asks the PerfectMatch algorithm for a car (line 59). This is ac-
tually a hard-coded combination of two algorithms. Now, what if the fallback algorithm of
PromotingMatch shall be changed? In this case it will be necessary to change the code on
line 59 in PromotingMatch. And what if a new combining algorithm is introduced? Say
for example DiscountMatch, which tries to give the customer a good deal by finding the car
with the highest discount, and if no discount is found uses another algorithm. Maybe it shall
be possible to change the fallback algorithm of both DiscountMatch and PromotingMatch
at runtime. And maybe DiscountMatch uses PromotingMatch as fallback, or vice versa.
All these situations prove it is a very bad solution to hard-code combinations of algorithms, as
was done in PromotingMatch. Instead, a composite must be used.

Solution

Since all concrete algorithms shall be independent of each other, the hard-coded fallback al-
gorithm in PromotingMatch is removed. All that is needed for this is to change line 59 in
listing 9.24, to return null if no car was found. This indicates that the algorithm did not find
a match.
The next step is to introduce a composite, which can be seen in listing 9.31. Note that this
class implements the same interface, Matcher, as the other matching algorithms. However,
no car matching is performed. Instead, match, which is supposed to search for cars, calls all
wrapped algorithms. The results can be combined in any way. Here, the search is interrupted
when an algorithm finds a car, and that car is then returned.
What about the empty init method on lines 53-55? This is the init method dis-
cussed above, in the section Try to make the factory independent of concrete products, on
page 210. The purpose is to completely remove all hard-coded class names of Matcher
implementations from the entire code. The addition of this method makes it possible to

215
 Chapter 9 Polymorphism and Inheritance

initialize PromotingMatch with the registration number of the car to promote, without
mentioning the name of that class in MatcherFactory. Exactly how this is done can
be seen in, se.leiflindback.oodbook.rentcarWithExAndDesPat.integration.
matchingWithComposite.MatcherFactory, in the accompanying GitHub repository
[Code].

1 /**
2 * A <code>Matcher</code>, which performs multiple matching
3 * algorithms. All matching algorithms added to this
4 * composite are executed, in the same order they were added,
5 * until an algorithm finds a car. Execution is stopped when
6 * an algorithm returns a non-null value.
7 */
8 class CompositeMatcher implements Matcher {
9 private List<Matcher> matchingAlgorithms =
10 new ArrayList<>();
11
12 CompositeMatcher() {
13 }
14
15 /**
16 * Invokes all matching algorithms added to this
17 * composite, in the same order they were added,
18 * until an algorithm finds a car. When a matching
19 * algorithm has found a car, that car is returned, and
20 * no more algorithms are called.
21 *
22 * @param searched Search criteria
23 * @param available Available cars
24 * @return A matching car, or <code>null</code> if none
25 * was found.
26 */
27 @Override
28 public CarDTO match(CarDTO searched,
29 List<CarDTO> available) {
30 for (Matcher matcher : matchingAlgorithms) {
31 CarDTO found = matcher.match(searched,
32 available);
33 if (found != null) {
34 return found;
35 }
36 }
37 return null;
38 }
39
40 /**

216
 Chapter 9 Polymorphism and Inheritance

41 * Adds a matching algorithm that will be invoked when

42 * this composite is searching for a car. The newly added
43 * algorithm will be called after all previously added
44 * algorithms, provided non of the previous algorithms
45 * finds a matching car.
46 *
47 * @param matcher The new <code>Matcher</code> to add.
48 */
49 void addMatcher(Matcher matcher) {
50 matchingAlgorithms.add(matcher);
51 }
52
53 @Override
54 public void init(Map<String, String> properties) {
55 }
56 }

Listing 9.31 CompositeMatcher wraps the concrete matchers which are passed to
addMatcher on lines 49-51. The match method calls these concrete matchers in the order they
were added, until a matcher finds a car.

How is the composite populated with Matchers? This is done by the fac-
tory. When it is asked to produce a concrete product, it reads the system property
se.leiflindback.rentcar.matcher.classname, just as previously. The difference
now, is that this property may contain not just one class name of a concrete matcher, but a
comma-separated list of such names. If more than one class is mentioned in the property, the
factory creates a CompositeMatcher, and adds all specified classes to it. The private method
responsible for creating the composite can be seen in listing 9.32.

1 private Matcher createComposite(String[] classNames) {

2 CompositeMatcher composite = new CompositeMatcher();
3 for (String className : classNames) {
4 composite.addMatcher(instantiateMatcher(className));
5 }
6 return composite;
7 }

Listing 9.32 Creation of composite matcher in MatcherFactory (exception handling is omit-

ted).

Comments

• But, there is still a hard-coded class name. Yes, CompositeMatcher is

mentioned on line two in listing 9.32. The question is if that class is consid-
ered part of the configurable algorithms, or if it is part of the infrastructure
supporting those. In the previous case, it should not be mentioned, whereas

217
 Chapter 9 Polymorphism and Inheritance

it might very well be mentioned in the latter case. It can be argued that
CompositeMatcher is, in fact, part of the infrastructure in the code just
written. There is only one composite, and it is used exactly the same way
whenever algorithms are to be combined.

• How to change the algorithm for combining outcomes of concrete algo-

rithms wrapped by a composite? Currently, CompositeMatcher always
returns the car found by the first algorithm that did find a car. Maybe this
way of choosing a result must be changeable. It might be desirable to have
the composite collect all cars found by all its algorithms, and then choose for
example the cheapest or the most expensive, depending on whether customer
satisfaction or short-term income shall be maximized. In this situation, there
must be one composite Matcher implementation for each desired way to
combine outcomes of matching algorithms. All that is needed to achieve
this, is to change a few lines in MatcherFactory. Instead of always instan-
tiating the same composite, the factory now must read also the class name
of the composite from the system property, or whatever method is used to
convey that information to the factory.

The Template Method Pattern

It is sometimes the case that the code in one class
is very similar to that in another class, except in
some detail. The principal strategy to remove the
duplicated code is to introduce a new method,
which contains the common code and can be
called by both classes. Unfortunately, this strat-
egy does not work if the duplicated code forms
the structure of a method. The duplicated code
might for example be a try-catch block or an
if statement, while the content of the blocks dif-
fer. In this situation, the template method pattern
is needed. Figure 9.23 A template leaves out only the parts
specific for a particular situation. Image by un-
known creator [Public domain], via https://pixabay.com

Overview

This pattern tells us to create a template, which, just like the invitation card in fig-
ure 9.23, contains everything common for all situations, and leaves out only the specific
parts. The template will be an abstract class, TaskTemplate in figure 9.24a, containing
a method, performTask, with the common code. This method calls an abstract method,
doPerformTask, when it is time to execute the part of the code that differs between con-
crete implementations of the task. This call will arrive at one of the concrete subclasses,
ConcreteTaskA in the example in figure 9.24b, which has overridden the abstract method
and provided an implementation.

218
 Chapter 9 Polymorphism and Inheritance

(a) Code that is common for performing the (b) The performTask method in the ab-
task of the two concrete classes, is placed stract template class calls a concrete imple-
in performTask, in the abstract template mentation of doPerformTask, when it
class. is time to execute code that differs between
concrete implementations.

Figure 9.24 Template method class and sequence diagrams.

Case Study

The case study concerns the display showing the number of rented cars, which was created as
a case study of the observer pattern. This display shows how many cars of different types have
currently been rented. The program shall now be extended to include two such displays, one
with a text view and one with a GUI view.

Solution

The source code of the display class can be seen in listing 9.18. The overridden method from
the RentalObserver interface, newRental on lines 20-24, calls two private methods. The
first, addNewRental, stores information about the latest rental, and is identical for the text and
GUI views. The second, printCurrentState, is different in the two views. This means that
newRental can be the same in both views, and also addNewRental. printCurrentState,
on the other hand, must be promoted to protected visibility, and overridden in the two
views. Listing 9.33 show the abstract template class, RentedCarsDisplay. Note the abstract
protected method on lines 38-39. Listings 9.34 and 9.35 show the two concrete views. They
contain only the implementations of printCurrentState, all other code is in the template
superclass. Finally, listing 9.36 shows part of the main view, View, where the concrete classes
are created.

1 /**
2 * Shows a running total of rented cars of each type.
3 */
4 public abstract class RentedCarsDisplay implements
5 RentalObserver {
6 private Map<CarDTO.CarType, Integer> noOfRentedCars =
7 new HashMap<>();
8
9 /**

219
 Chapter 9 Polymorphism and Inheritance

10 * Creates a new instance, with the all counters of rented

11 * cars set to zero.
12 */
13 protected RentedCarsDisplay() {
14 for (CarDTO.CarType type : CarDTO.CarType.values()) {
15 noOfRentedCars.put(type, 0);
16 }
17 }
18
19 @Override
20 public void newRental(CarDTO rentedCar) {
21 addNewRental(rentedCar);
22 printCurrentState(noOfRentedCars);
23 }
24
25 private void addNewRental(CarDTO rentedCar) {
26 int noOfRentedCarsOfThisType =
27 noOfRentedCars.get(rentedCar.getSize()) + 1;
28 noOfRentedCars.put(rentedCar.getSize(),
29 noOfRentedCarsOfThisType);
30 }
31
32 /**
33 * Shows the number of rented cars.
34 *
35 * @param noOfRentedCars Contains the number of rented
36 * cars of each type.
37 */
38 protected abstract void printCurrentState(
39 Map<CarDTO.CarType, Integer> noOfRentedCars);
40 }

Listing 9.33 The template class, which leaves the adaptable parts in an abstract protected method,
lines 38-39

1 /**
2 * Prints the number of rented cars on the console.
3 */
4 public class ConsoleRentedCarsDisplay extends
5 RentedCarsDisplay {
6 /**
7 * Prints the number of rented cars of each type on the
8 * console.
9 */
10 @Override
11 protected void printCurrentState(

220
 Chapter 9 Polymorphism and Inheritance

12 Map<CarDTO.CarType, Integer> noOfRentedCars) {

13 System.out.println("### We have now rented out ###");
14 for (CarDTO.CarType type : CarDTO.CarType.values()) {
15 System.out.print(noOfRentedCars.get(type));
16 System.out.print(" ");
17 System.out.print(type.toString().toLowerCase());
18 System.out.println(" cars.");
19 }
20 System.out.println("##############################");
21 }
22 }
Listing 9.34 A concrete implementation of the template. It contains only code that is specific for
this particular implemenation

1 /**
2 * Shows a GUI with the number of rented cars.
3 */
4 public class GuiRentedCarsDisplay extends RentedCarsDisplay {
5 private Display display;
6
7 /**
8 * Starts the GUI.
9 */
10 public GuiRentedCarsDisplay() {
11 new Thread(() -> {
12 Application.launch(Display.class, null);
13 }).start();
14 this.display = Display.getDisplay();
15 }
16
17 /**
18 * Shows a GUI with the number of rented cars of
19 * each type.
20 */
21 @Override
22 protected void printCurrentState(
23 Map<CarDTO.CarType, Integer> noOfRentedCars) {
24 display.updateStatsList(noOfRentedCars);
25 }
26
27 // Code for GUI handling.
28 }
Listing 9.35 A concrete implementation of the template. It contains only code that is specific for
this particular implemenation

221
 Chapter 9 Polymorphism and Inheritance

1 /**
2 * This program has no view, instead, this class is a
3 * placeholder for the entire view.
4 */
5 public class View {
6 private Controller contr;
7 private ErrorMessageHandler errorMsgHandler =
8 new ErrorMessageHandler();
9 private LogHandler logger = LogHandler.getLogger();
10
11 /**
12 * Creates a new instance.
13 *
14 * @param contr The controller that is used for all
15 * operations.
16 */
17 public View(Controller contr) throws IOException {
18 this.contr = contr;
19 contr.addRentalObserver(
20 new ConsoleRentedCarsDisplay());
21 contr.addRentalObserver(new GuiRentedCarsDisplay());
22 }
23
24 // More methods, not relevant for this listing.
25 }

Listing 9.36 Creating the concerete implementations, lines 20 and 21

Comments

• More abstract methods In all examples mentioned here, there was only one
abstract method in the template class. This is not a requirement, we are free
to add any number of abstract methods, consider for example the template
class delineated in listing 9.37

1 /**
2 * An example of a template class with more than one abstract
3 * method.
4 */
5 public abstract class TemplateWithMoreAbstractMethods {
6 public void theTemplateMethod() {
7 try {
8 performSomeTask();
9 } catch (Exception e) {

222
 Chapter 9 Polymorphism and Inheritance

10 errorHandling();
11 } finally {
12 cleanup();
13 }
14 }
15
16 protected abstract void performSomeTask();
17
18 protected abstract void errorHandling();
19
20 protected abstract void cleanup();
21 }

Listing 9.37 A template class with more than one abstract method

223
 Part III

Appendices

224
Appendix A

English-Swedish Dictionary
This appendix contains translations to Swedish of some English terms in the text. Be aware
that these translations are terms commonly used in software development, and have a defined
meaning. To know a general translation because of skills in English language is not enough,
exactly the correct term must be used, not a synonym. Having said that, it is also important to
point out that there is no universal agreement on these terms, do not be surprised when finding
other words meaning the same thing.

English Swedish
analysis . . . . . . . . . . . . . . . . . . . analys
architectural pattern . . . . . . . . arkitekturellt mönster
architecture . . . . . . . . . . . . . . . arkitektur
class diagram . . . . . . . . . . . . . . klassdiagram
code convention . . . . . . . . . . . kodkonvention
communication diagram . . . . kommunikationsdiagram
design . . . . . . . . . . . . . . . . . . . . design
design pattern . . . . . . . . . . . . . designmönster
domain model . . . . . . . . . . . . . domänmodell
encapsulation . . . . . . . . . . . . . . inkapsling
entity . . . . . . . . . . . . . . . . . . . . . entitet
enumeration . . . . . . . . . . . . . . . uppräkningsbar typ
high cohesion . . . . . . . . . . . . . hög sammanhållning
implement . . . . . . . . . . . . . . . . implementera
instance . . . . . . . . . . . . . . . . . . . instans
layer . . . . . . . . . . . . . . . . . . . . . . lager
low coupling . . . . . . . . . . . . . . låg koppling
modifier . . . . . . . . . . . . . . . . . . modifierare
overload . . . . . . . . . . . . . . . . . . överlagra
override . . . . . . . . . . . . . . . . . . . omdefiniera
package diagram . . . . . . . . . . . paketdiagram
pattern . . . . . . . . . . . . . . . . . . . . mönster

225
 Appendix A English-Swedish Dictionary

English Swedish
refactoring . . . . . . . . . . . . . . . . omstrukturering, refaktorering, refaktorisering
sequence diagram . . . . . . . . . . sekvensdiagram
system operation . . . . . . . . . . . systemoperation
system sequence diagram . . . systemsekvensdiagram
visibility . . . . . . . . . . . . . . . . . . åtkomst

226
Appendix B

UML Cheat Sheet

Class Diagram

Figure B.1 Class diagram

227
 Appendix B UML Cheat Sheet

Figure B.2 Exceptions and package names in class diagram

Figure B.3 Packages in class diagram

228
 Appendix B UML Cheat Sheet

Sequence Diagram

Figure B.4 Sequence diagram

Figure B.5 Flow control in sequence diagram

229
 Appendix B UML Cheat Sheet

Figure B.6 Reference to other sequence diagram

Communication Diagram

Figure B.7 Communication diagram

230
 Appendix B UML Cheat Sheet

Figure B.8 Flow control in communication diagram

Figure B.9 Exception handling in communication diagram

231
Appendix C

Implementations of UML Diagrams

This appendix contains Java implementations of all UML design diagrams in the text. The
purpose is to make clearer what the diagrams actually mean. The analysis diagrams can not
be implemented in code, since they do not represent programs.

C.1 Figure 5.1

Package names are not shown in the diagram, but have been added in the code.

1 package se.kth.ict.oodbook.design.uml;
2
3 public class AClass {
4 public void aMethod(int aParam) {
5 }
6 }

Listing C.1 Java code implementing the AClass

class in figure 5.1a

1 package se.kth.ict.oodbook.design.uml;
2
3 public class AnotherClass {
4 private static int aStaticAttribute;
5
6 public static String aStaticMethod() {
7 }
8 }

Listing C.2 Java code implementing the AnotherClass

class in figure 5.1b

1 package se.kth.ict.oodbook.design.uml;
2
3 public class YetAnotherClass {
4 private int privateAttribute;
5 public int publicAttribute;

232
 Appendix C Implementations of UML Diagrams

6
7 private String privateMethod() {
8 }
9
10 public int publicMethod() {
11 }
12 }

Listing C.3 Java code implementing the

YetAnotherClass class in figure 5.1c

C.2 Figure 5.2

The diagram tells somePackage in some way depends on someOtherPackage, but that can
not be implemented in code since it does not tell how.

1 package somePackage;

Listing C.4
Java code im-
plementing the
somePackage
package in figure
5.2

1 package someOtherPackage;

Listing C.5 Java code

implementing the
someOtherPackage
package in figure 5.2

C.3 Figure 5.3

Package names are not shown in the diagram, but have been added in the code. Visibility is
also not shown in the diagram. Here, attributes and methods called by objects where they
are located have been assigned private visibility, while constructors and other methods have
public visibility.

1 package se.kth.ict.oodbook.design.uml;
2
3 public class SomeClass {
4 private OtherClass otherObj;
5
6 public void firstMethod() {

233
 Appendix C Implementations of UML Diagrams

7 otherObj.aMethod();
8 methodInSelf();
9 }
10
11 public void someMethod() {
12 }
13
14 private void methodInSelf() {
15 }
16 }

Listing C.6 Java code implementing the SomeClass

class in figure 5.3

1 package se.kth.ict.oodbook.design.uml;
2
3 public class OtherClass {
4 private SomeClass someObj;
5
6 public void aMethod() {
7 someObj.someMethod();
8 ThirdClass newObj = new ThirdClass();
9 }
10 }

Listing C.7 Java code implementing the OtherClass class in fig-

ure 5.3

1 package se.kth.ict.oodbook.design.uml;
2
3 public class ThirdClass {
4 public ThirdClass() {
5 }
6 }

Listing C.8 Java code implementing the

ThirdClass class in figure 5.3

C.4 Figure 5.4

Package names are not shown in the diagram, but have been added in the code. Visibility is
also not shown in the diagram. Here, attributes and methods called by objects where they are
located have been assigned private visibility, while other methods have public visibility.

1 package se.kth.ict.oodbook.design.uml;

234
 Appendix C Implementations of UML Diagrams

2
3 public class A {
4 private B b;
5 // Somewhere in some method the following call is made:
6 // b.met1();
7 }

Listing C.9 Java code implementing the A class in figure 5.4

1 package se.kth.ict.oodbook.design.uml;
2
3 public class B {
4 public void met1() {
5 C.met2();
6 }
7 /*
8 * Code illustrated in a sequence diagram named ’SomeTask’.
9 */
10 }

Listing C.10 Java code implementing the B class in figure 5.4

1 package se.kth.ict.oodbook.design.uml;
2
3 public class C {
4 public static void met2() {
5
6 }
7 /*
8 * Code illustrated in a sequence diagram named ’SomeTask’.
9 */
10 }

Listing C.11 Java code implementing the C class in figure 5.4

C.5 Figure 5.5

Package names are not shown in the diagram, but have been added in the code. Visibility is
also not shown in the diagram. Here, attributes and methods called by objects where they
are located have been assigned private visibility, while constructors and methods have public
visibility.

1 package se.kth.ict.oodbook.design.uml;

235
 Appendix C Implementations of UML Diagrams

2
3 public class ClassA {
4 private ClassB objB;
5 private ClassE objE;
6
7 public void metF() {
8 }
9 /* The following lines appear somwhere in the code, in the
10 * order they are written here.
11 * objB.metA(2);
12 * int retVal = objE.metD();
13 * objE.metE();
14 */
15 }

Listing C.12 Java code implementing the ClassA class in figure 5.5

1 package se.kth.ict.oodbook.design.uml;
2
3 public class ClassB {
4 private ClassC objC;
5
6 public void metA(int aParam) {
7 objC.metB();
8 ClassD objD = new ClassD();
9 }
10 }

Listing C.13 Java code implementing the ClassB

class in figure 5.5

1 package se.kth.ict.oodbook.design.uml;
2
3 public class ClassC {
4 public void metB() {
5 }
6 }

Listing C.14 Java code implementing the ClassC

class in figure 5.5

1 package se.kth.ict.oodbook.design.uml;
2
3 public class ClassD {
4 public ClassD() {
5 myMethod();

236
 Appendix C Implementations of UML Diagrams

6 }
7
8 private void myMethod() {
9 }
10 }

Listing C.15 Java code implementing the ClassD

class in figure 5.5

237
 Appendix C Implementations of UML Diagrams

1 package se.kth.ict.oodbook.design.uml;
2
3 public class ClassE {
4 private ClassA objA;
5
6 public void metE() {
7 objA.metF();
8 }
9
10 public int metD() {
11 return 0;
12 }
13 }

Listing C.16 Java code implementing the ClassE

class in figure 5.5

C.6 Figure 5.6

Package names are not shown in the diagram, but have been added in the code. Visibility is
also not shown in the diagram. Here, attributes and methods called by objects where they
are located have been assigned private visibility, while constructors and methods have public
visibility.

1 package se.kth.ict.oodbook.design.uml;
2
3 public class ClassF {
4 private int count;
5 private ClassG classG;
6 private ClassH classH;
7
8 /* The following code appears somewhere, in some method:
9 * if (count == 3) {
10 * classG.aMethod();
11 * } else {
12 * classH.aMethod();
13 * }
14 */
15 }

Listing C.17 Java code implementing the ClassF class in figure 5.6

1 package se.kth.ict.oodbook.design.uml;
2

238
 Appendix C Implementations of UML Diagrams

3 public class ClassG {

4 public void aMethod() {
5 }
6 }

Listing C.18 Java code implementing the ClassG

class in figure 5.6

1 package se.kth.ict.oodbook.design.uml;
2
3 public class ClassH {
4 private int n;
5 private ClassK classK;
6
7 public void aMethod() {
8 for (int i = 1; i <= n; i++) {
9 classK.aMet();
10 }
11 }
12 }

Listing C.19 Java code implementing the ClassH

class in figure 5.6

1 package se.kth.ict.oodbook.design.uml;
2
3 public class ClassK {
4 public void aMet() {
5 }
6 }

Listing C.20 Java code implementing the ClassK

class in figure 5.6

C.7 Figure 5.9

The method bodies on line 17 in listing C.22 and line 17 in listing C.23 are not shown in the
diagram in figure 5.9b. The code on those lines is included here since there is no reasonable
alternative. The attribute on line eleven in listing C.23 is illustrated by the association in figure
5.9b. Package names are not shown in the diagram, but has been added in the code.

1 package se.kth.ict.oodbook.design.cohesion;
2
3 import java.util.List;
4

239
 Appendix C Implementations of UML Diagrams

5 /**
6 * Represents an employee.
7 */
8 public class BadDesignEmployee {
9 private String name;
10 private Address address;
11 private Amount salary;
12
13 /**
14 * Changes the salary of <code>employee</code> to
15 * <code>newSalary</code>.
16 *
17 * @param employee The <code>Employee</code> whose salary will be
18 * changed.
19 * @param newSalary The new salary of <code>employee</code>.
20 */
21 public void changeSalary(BadDesignEmployee employee,
22 Amount newSalary) {
23 }
24
25 /**
26 * Returns a list with all employees working in the same
27 * department as this employee.
28 */
29 public List<BadDesignEmployee> getAllEmployees() {
30 }
31 }

Listing C.21 Java code implementing the UML diagram in figure 5.9a

1 package se.kth.ict.oodbook.design.cohesion;
2
3 /**
4 * Represents an employee.
5 */
6 public class Employee {
7 private String name;
8 private Address address;
9 private Amount salary;
10
11 /**
12 * Changes the salary to <code>newSalary</code>.
13 *
14 * @param newSalary The new salary.
15 */
16 public void changeSalary(Amount newSalary) {

240
 Appendix C Implementations of UML Diagrams

17 this.salary = newSalary;
18 }
19 }

Listing C.22 Java code implementing the Employee class in figure 5.9b

1 package se.kth.ict.oodbook.design.cohesion;
2
3 import java.util.ArrayList;
4 import java.util.List;
5
6 /**
7 * Represents a department.
8 */
9 public class Department {
10 private String name;
11 private List<Employee> employees = new ArrayList<>();
12
13 /**
14 * Returns a list with all employees working in this department.
15 */
16 public List<Employee> getEmployees() {
17 return employees;
18 }
19 }

Listing C.23 Java code implementing the Department class in figure 5.9b

C.8 Figure 5.10

The method body on line 15 in listing C.24 is not shown in the diagram in figure 5.10a. The
code on that line is included here since there is no reasonable alternative.
The method body on line 16 in listing C.25 is not shown in the diagram in figure 5.10b. The
code on that line is included here since there is no reasonable alternative. The attributes on
lines nine and ten in listing C.25 are illustrated by the associations in figure 5.10b. Package
names are not shown in the diagram, but has been added in the code.

1 package se.kth.ict.oodbook.design.cohesion;
2
3 /**
4 * Represents a car.
5 */
6 public class BadDesignCar {
7 private String regNo;

241
 Appendix C Implementations of UML Diagrams

8 private Person owner;

9 private String ownersPreferredRadioStation;
10
11 /**
12 * Returns the registration number of this car.
13 */
14 public String getRegNo() {
15 return regNo;
16 }
17
18 /**
19 * Accelerates the car.
20 */
21 public void accelerate() {
22
23 }
24
25 /**
26 * Breaks the car.
27 */
28 public void brake() {
29
30 }
31
32 /**
33 * Sets the car’s radio to the specified station.
34 * @param station The station to which to listen.
35 */
36 public void changeRadioStation(String station) {
37
38 }
39 }

Listing C.24 Java code implementing the BadDesignCar class in figure 5.10a

1 package se.kth.ict.oodbook.design.cohesion;
2
3 /**
4 * Represents a car.
5 */
6 public class Car {
7 private String regNo;
8 private Person owner;
9 private Engine engine;
10 private Radio radio;
11

242
 Appendix C Implementations of UML Diagrams

12 /**
13 * Returns the registration number of this car.
14 */
15 public String getRegNo() {
16 return regNo;
17 }
18 }

Listing C.25 Java code implementing the Car class in figure 5.10b

1 package se.kth.ict.oodbook.design.cohesion;
2
3 /**
4 * Represents a car radio.
5 */
6 public class Radio {
7 private String ownersPreferredStation;
8 /**
9 * Sets the radio to the specified station.
10 * @param station The station to which to listen.
11 */
12 public void changeStation(String station) {
13 }
14 }

Listing C.26 Java code implementing the Radio class in figure 5.10b

1 package se.kth.ict.oodbook.design.cohesion;
2
3 /**
4 * Represents a car engine.
5 */
6 public class Engine {
7 /**
8 * Accelerates the car.
9 */
10 public void accelerate() {
11 }
12
13 /**
14 * Breaks the car.
15 */
16 public void brake() {
17 }
18 }

243
 Appendix C Implementations of UML Diagrams

Listing C.27 Java code implementing the Engine class in figure

5.10b

C.9 Figure 5.12

Package names are not shown in the diagram, but have been added in the code.

1 package se.kth.ict.oodbook.design.coupling;
2
3 public class HighCouplingOrder {
4 private HighCouplingCustomer customer;
5 private HighCouplingShippingAddress shippingAddress;
6 }

Listing C.28 Java code implementing the HighCouplingOrder class in figure

5.12a

1 package se.kth.ict.oodbook.design.coupling;
2
3 class HighCouplingCustomer {
4 private HighCouplingShippingAddress shippingAddress;
5 }

Listing C.29 Java code implementing the HighCouplingCustomer class in figure

5.12a

1 package se.kth.ict.oodbook.design.coupling;
2
3 class HighCouplingShippingAddress {
4 }

Listing C.30 Java code implementing the HighCouplingShippingAddress

class in figure 5.12a

1 package se.kth.ict.oodbook.design.coupling;
2
3 public class Order {
4 private Customer customer;
5 }

Listing C.31 Java code implementing the Order class in figure 5.12b

1 package se.kth.ict.oodbook.design.coupling;

244
 Appendix C Implementations of UML Diagrams

2
3 class Customer {
4 private ShippingAddress shippingAddress;
5 }

Listing C.32 Java code implementing the Customer class in figure 5.12b

1 package se.kth.ict.oodbook.design.coupling;
2
3 class ShippingAddress {
4 }

Listing C.33 Java code implementing the ShippingAddress class in figure 5.12b

C.10 Figure 5.13

Package names are not shown in the diagram, but have been added in the code.

1 package se.kth.ict.oodbook.design.coupling;
2
3 class HighCouplingBooking {
4 }

Listing C.34 Java code implementing the HighCouplingBooking class in figure

5.13a

1 package se.kth.ict.oodbook.design.coupling;
2
3 class HighCouplingGuest {
4 }

Listing C.35 Java code implementing the HighCouplingGuest class in figure

5.13a

1 package se.kth.ict.oodbook.design.coupling;
2
3 public class HighCouplingHotel {
4 private HighCouplingBooking booking;
5 private HighCouplingGuest guest;
6 private HighCouplingAddress address;
7 private HighCouplingFloor floor;
8 private HighCouplingRoom room;
9 }

245
 Appendix C Implementations of UML Diagrams

Listing C.36 Java code implementing the HighCouplingHotel class in figure

5.13a

246
 Appendix C Implementations of UML Diagrams

1 package se.kth.ict.oodbook.design.coupling;
2
3 class HighCouplingAddress {
4 }

Listing C.37 Java code implementing the HighCouplingAddress class in figure

5.13a

1 package se.kth.ict.oodbook.design.coupling;
2
3 class HighCouplingFloor {
4 }

Listing C.38 Java code implementing the HighCouplingFloor class in figure

5.13a

1 package se.kth.ict.oodbook.design.coupling;
2
3 class HighCouplingRoom {
4 }

Listing C.39 Java code implementing the HighCouplingRoom class in figure 5.13a

1 package se.kth.ict.oodbook.design.coupling;
2
3 class Booking {
4 private Guest guest;
5 }

Listing C.40 Java code implementing the Booking class in figure 5.13b

1 package se.kth.ict.oodbook.design.coupling;
2
3 class Guest {
4 }

Listing C.41 Java code implementing the Guest class in figure 5.13b

1 package se.kth.ict.oodbook.design.coupling;
2
3 public class Hotel {
4 private Booking booking;
5 private Address address;

247
 Appendix C Implementations of UML Diagrams

6 private Floor floor;

7 }

Listing C.42 Java code implementing the Hotel class in figure 5.13b

1 package se.kth.ict.oodbook.design.coupling;
2
3 class Address {
4 }

Listing C.43 Java code implementing the Address class in figure 5.13b

1 package se.kth.ict.oodbook.design.coupling;
2
3 class Floor {
4 private Room room;
5 }

Listing C.44 Java code implementing the Floor class in figure 5.13b

1 package se.kth.ict.oodbook.design.coupling;
2
3 class Room {
4 }

Listing C.45 Java code implementing the Room class in figure 5.13b

248
 Appendix C Implementations of UML Diagrams

C.11 Figure 5.15

Package name is not shown in the diagram, but has been added in the code.

1 package se.kth.ict.oodbook.architecture.packPriv;
2
3 /**
4 * Illustrates package private field and method. Note that it is
5 * not required to write javadoc for these, since they are not
6 * part of the public interface.
7 */
8 public class PackPriv {
9 int packagePrivateAtribute;
10
11 void packagePrivateMethod() {
12 }
13 }

Listing C.46 Java code implementing the PackPriv class in figure 5.15

C.12 Figure 5.18

1 package se.kth.ict.oodbook.architecture.mvc.controller;
2
3 /**
4 * This is the application’s controller. All calls from view to model
5 * pass through here.
6 */
7 public class Controller {
8 /**
9 * A system operation, which means it appears in the system sequence
10 * diagram.
11 */
12 public void systemOperation1() {
13 }
14
15 /**
16 * A system operation, which means it appears in the system sequence
17 * diagram.
18 */
19 public void systemOperation2() {
20 }
21 }

Listing C.47 Java code implementing the Controller class in figure 5.18

249
 Appendix C Implementations of UML Diagrams

C.13 Figure 5.19

1 package se.kth.ict.oodbook.architecture.mvc.view;
2
3 import se.kth.ict.oodbook.architecture.mvc.controller.Controller;
4
5 /**
6 * A class in the view.
7 */
8 public class ClassInView {
9 private Controller contr;
10
11 //Somewhere in some method.
12 contr.systemOperation1();
13 }

Listing C.48 Java code implementing the ClassInView class in figure 5.19

1 package se.kth.ict.oodbook.architecture.mvc.controller;
2
3 import se.kth.ict.oodbook.architecture.mvc.model.OtherClassInModel;
4 import se.kth.ict.oodbook.architecture.mvc.model.SomeClassInModel;
5
6 /**
7 * This is the application’s controller. All calls from view to model
8 * pass through here.
9 */
10 public class Controller {
11 private SomeClassInModel scim;
12 private OtherClassInModel ocim;
13
14 /**
15 * A system operation, which means it appears in the system sequence
16 * diagram.
17 */
18 public void systemOperation1() {
19 scim.aMethod();
20 ocim.aMethod();
21 }
22 }

Listing C.49 Java code implementing the Controller class in figure 5.19

250
 Appendix C Implementations of UML Diagrams

1 package se.kth.ict.oodbook.architecture.mvc.model;
2
3 /**
4 * A class in the model, performing some business logic.
5 */
6 public class SomeClassInModel {
7
8 /**
9 * Performs some business logic.
10 */
11 public void aMethod() {
12 }
13 }

Listing C.50 Java code implementing the SomeClassInModel class in figure 5.19

1 package se.kth.ict.oodbook.architecture.mvc.model;
2
3 /**
4 * A class in the model, performing some business logic.
5 */
6 public class OtherClassInModel {
7
8 /**
9 * Performs some business logic.
10 */
11 public void aMethod() {
12 }
13 }

Listing C.51 Java code implementing the OtherClassInModel class in figure 5.19

C.14 Figure 5.25

1 package se.kth.ict.oodbook.design.casestudy.view;
2
3 import se.kth.ict.oodbook.design.casestudy.controller.Controller;
4 import se.kth.ict.oodbook.design.casestudy.dbhandler.Car;
5
6 /**
7 * This program has no view, instead, this class is a placeholder
8 * for the entire view.

251
 Appendix C Implementations of UML Diagrams

9 */
10 public class View {
11 // Somewhere in the code. Note that the arguments to the
12 // Car constructor are not specified in the UML diagram.
13 Car searchedCar = new Car(0, null, false, false,
14 null, null);
15 Car foundCar = contr.searchMatchingCar(searchedCar);
16 }
17
18 }

Listing C.52 Java code implementing the View class in figure 5.25

1 package se.kth.ict.oodbook.design.casestudy.controller;
2
3 import se.kth.ict.oodbook.design.casestudy.dbhandler.Car;
4
5 /**
6 * This is the application’s only controller class. All calls to
7 * the model pass through here.
8 */
9 public class Controller {
10 public Car searchMatchingCar(Car searchedCar) {
11 }
12 }

Listing C.53 Java code implementing the Controller class in figure 5.25

1 package se.kth.ict.oodbook.design.casestudy.dbhandler;
2
3 /**
4 * Contains information about one particular car.
5 */
6 public class Car {
7
8 private int price;
9 private String size;
10 private boolean AC;
11 private boolean fourWD;
12 private String color;
13 private String regNo;
14
15 /**
16 * Creates a new instance representing a particular car.
17 *

252
 Appendix C Implementations of UML Diagrams

18 * @param price The price paid to rent the car.

19 * @param size The size of the car, e.g., <code>medium
20 hatchback</code>.
21 * @param AC <code>true</code> if the car has air
22 condition.
23 * @param fourWD <code>true</code> if the car has four
24 wheel drive.
25 * @param color The color of the car.
26 * @param regNo The car’s registration number.
27 */
28 public Car(int price, String size, boolean AC, boolean fourWD,
29 String color, String regNo) {
30 this.price = price;
31 this.size = size;
32 this.AC = AC;
33 this.fourWD = fourWD;
34 this.color = color;
35 this.regNo = regNo;
36 }
37
38 /**
39 * Get the value of regNo
40 *
41 * @return the value of regNo
42 */
43 public String getRegNo() {
44 return regNo;
45 }
46
47 /**
48 * Get the value of color
49 *
50 * @return the value of color
51 */
52 public String getColor() {
53 return color;
54 }
55
56 /**
57 * Get the value of fourWD
58 *
59 * @return the value of fourWD
60 */
61 public boolean isFourWD() {
62 return fourWD;
63 }

253
 Appendix C Implementations of UML Diagrams

64
65 /**
66 * Get the value of AC
67 *
68 * @return the value of AC
69 */
70 public boolean isAC() {
71 return AC;
72 }
73
74 /**
75 * Get the value of size
76 *
77 * @return the value of size
78 */
79 public String getSize() {
80 return size;
81 }
82
83 /**
84 * Get the value of price
85 *
86 * @return the value of price
87 */
88 public int getPrice() {
89 return price;
90 }
91
92 }

Listing C.54 Java code implementing the Car class in figure 5.25

C.15 Figure 5.26

1 package se.kth.ict.oodbook.design.casestudy.view;
2
3 import se.kth.ict.oodbook.design.casestudy.controller.Controller;
4 import se.kth.ict.oodbook.design.casestudy.dbhandler.CarDTO;
5
6 /**
7 * This program has no view, instead, this class is a placeholder
8 * for the entire view.
9 */
10 public class View {

254
 Appendix C Implementations of UML Diagrams

11 // Somewhere in the code. Note that the arguments to the

12 // CarDTO constructor are not specified in the UML
13 // diagram.
14 CarDTO searchedCar = new CarDTO(0, null, false, false,
15 null, null);
16 CarDTO foundCar = contr.searchMatchingCar(searchedCar);
17 }
18
19 }

Listing C.55 Java code implementing the View class in figure 5.26

1 package se.kth.ict.oodbook.design.casestudy.controller;
2
3 import se.kth.ict.oodbook.design.casestudy.dbhandler.CarDTO;
4
5 /**
6 * This is the application’s only controller class. All calls to
7 * the model pass through here.
8 */
9 public class Controller {
10 public CarDTO searchMatchingCar(CarDTO searchedCar) {
11 }
12 }

Listing C.56 Java code implementing the Controller class in figure 5.26

1 package se.kth.ict.oodbook.design.casestudy.dbhandler;
2
3 /**
4 * Contains information about one particular car.
5 */
6 public class CarDTO {
7
8 private int price;
9 private String size;
10 private boolean AC;
11 private boolean fourWD;
12 private String color;
13 private String regNo;
14
15 /**
16 * Creates a new instance representing a particular car.
17 *
18 * @param price The price paid to rent the car.

255
 Appendix C Implementations of UML Diagrams

19 * @param size The size of the car, e.g., <code>medium

20 hatchback</code>.
21 * @param AC <code>true</code> if the car has air
22 condition.
23 * @param fourWD <code>true</code> if the car has four
24 wheel drive.
25 * @param color The color of the car.
26 * @param regNo The car’s registration number.
27 */
28 public CarDTO(int price, String size, boolean AC,
29 boolean fourWD, String color, String regNo) {
30 this.price = price;
31 this.size = size;
32 this.AC = AC;
33 this.fourWD = fourWD;
34 this.color = color;
35 this.regNo = regNo;
36 }
37
38 /**
39 * Get the value of regNo
40 *
41 * @return the value of regNo
42 */
43 public String getRegNo() {
44 return regNo;
45 }
46
47 /**
48 * Get the value of color
49 *
50 * @return the value of color
51 */
52 public String getColor() {
53 return color;
54 }
55
56 /**
57 * Get the value of fourWD
58 *
59 * @return the value of fourWD
60 */
61 public boolean isFourWD() {
62 return fourWD;
63 }
64

256
 Appendix C Implementations of UML Diagrams

65 /**
66 * Get the value of AC
67 *
68 * @return the value of AC
69 */
70 public boolean isAC() {
71 return AC;
72 }
73
74 /**
75 * Get the value of size
76 *
77 * @return the value of size
78 */
79 public String getSize() {
80 return size;
81 }
82
83 /**
84 * Get the value of price
85 *
86 * @return the value of price
87 */
88 public int getPrice() {
89 return price;
90 }
91
92 }

Listing C.57 Java code implementing the CarDTO class in figure 5.26

C.16 Figure 5.27

1 package se.kth.ict.oodbook.design.casestudy.view;
2
3 import se.kth.ict.oodbook.design.casestudy.controller.Controller;
4 import se.kth.ict.oodbook.design.casestudy.dbhandler.CarDTO;
5
6 /**
7 * This program has no view, instead, this class is a placeholder
8 * for the entire view.
9 */
10 public class View {
11 // Somewhere in the code. Note that the arguments to the

257
 Appendix C Implementations of UML Diagrams

12 // CarDTO constructor are not specified in the UML

13 // diagram.
14 CarDTO searchedCar = new CarDTO(0, null, false, false,
15 null, null);
16 CarDTO foundCar = contr.searchMatchingCar(searchedCar);
17 }
18
19 }

Listing C.58 Java code implementing the View class in figure 5.27

1 package se.kth.ict.oodbook.design.casestudy.controller;
2
3 import se.kth.ict.oodbook.design.casestudy.dbhandler.CarDTO;
4
5 /**
6 * This is the application’s only controller class. All calls to
7 * the model pass through here.
8 */
9 public class Controller {
10 public CarDTO searchMatchingCar(CarDTO searchedCar) {
11 return carRegistry.findCar(searchedCar);
12 }
13 }

Listing C.59 Java code implementing the Controller class in figure 5.27

1 package se.kth.ict.oodbook.design.casestudy.dbhandler;
2
3 /**
4 * Contains information about one particular car.
5 */
6 public class CarDTO {
7
8 private int price;
9 private String size;
10 private boolean AC;
11 private boolean fourWD;
12 private String color;
13 private String regNo;
14
15 /**
16 * Creates a new instance representing a particular car.
17 *
18 * @param price The price paid to rent the car.

258
 Appendix C Implementations of UML Diagrams

19 * @param size The size of the car, e.g., <code>medium

20 hatchback</code>.
21 * @param AC <code>true</code> if the car has air
22 condition.
23 * @param fourWD <code>true</code> if the car has four
24 wheel drive.
25 * @param color The color of the car.
26 * @param regNo The car’s registration number.
27 */
28 public CarDTO(int price, String size, boolean AC,
29 boolean fourWD, String color, String regNo) {
30 this.price = price;
31 this.size = size;
32 this.AC = AC;
33 this.fourWD = fourWD;
34 this.color = color;
35 this.regNo = regNo;
36 }
37
38 /**
39 * Get the value of regNo
40 *
41 * @return the value of regNo
42 */
43 public String getRegNo() {
44 return regNo;
45 }
46
47 /**
48 * Get the value of color
49 *
50 * @return the value of color
51 */
52 public String getColor() {
53 return color;
54 }
55
56 /**
57 * Get the value of fourWD
58 *
59 * @return the value of fourWD
60 */
61 public boolean isFourWD() {
62 return fourWD;
63 }
64

259
 Appendix C Implementations of UML Diagrams

65 /**
66 * Get the value of AC
67 *
68 * @return the value of AC
69 */
70 public boolean isAC() {
71 return AC;
72 }
73
74 /**
75 * Get the value of size
76 *
77 * @return the value of size
78 */
79 public String getSize() {
80 return size;
81 }
82
83 /**
84 * Get the value of price
85 *
86 * @return the value of price
87 */
88 public int getPrice() {
89 return price;
90 }
91
92 }

Listing C.60 Java code implementing the CarDTO class in figure 5.27

1 package se.kth.ict.oodbook.design.casestudy.dbhandler;
2
3 /**
4 * Contains all calls to the data store with cars that may be
5 * rented.
6 */
7 public class CarRegistry {
8 public CarDTO findCar(CarDTO searchedCar) {
9 }
10 }

Listing C.61 Java code implementing the CarRegistry class in figure 5.27

260
 Appendix C Implementations of UML Diagrams

C.17 Figure 5.28

1 package se.kth.ict.oodbook.design.casestudy.startup;
2
3 import se.kth.ict.oodbook.design.casestudy.controller.Controller;
4 import se.kth.ict.oodbook.design.casestudy.dbhandler.CarRegistry;
5 import se.kth.ict.oodbook.design.casestudy.view.View;
6
7 /**
8 * Contains the <code>main</code> method. Performs all startup of
9 * the application.
10 */
11 public class Main {
12 public static void main(String[] args) {
13 CarRegistry carRegistry = new CarRegistry();
14 Controller contr = new Controller(carRegistry);
15 new View(contr);
16 }
17 }

Listing C.62 Java code implementing the Main class in figure 5.28

1 package se.kth.ict.oodbook.design.casestudy.dbhandler;
2
3 /**
4 * Contains all calls to the data store with cars that may be
5 * rented.
6 */
7 public class CarRegistry {
8 }

Listing C.63 Java code implementing the CarRegistry class in figure 5.28

1 package se.kth.ict.oodbook.design.casestudy.controller;
2
3 import se.kth.ict.oodbook.design.casestudy.dbhandler.CarDTO;
4 import se.kth.ict.oodbook.design.casestudy.dbhandler.CarRegistry;
5
6 /**
7 * This is the application’s only controller class. All calls to
8 * the model pass through here.
9 */
10 public class Controller {
11 public Controller(CarRegistry carRegistry) {

261
 Appendix C Implementations of UML Diagrams

12 }
13 }

Listing C.64 Java code implementing the Controller class in figure 5.28

1 package se.kth.ict.oodbook.design.casestudy.view;
2
3 import se.kth.ict.oodbook.design.casestudy.controller.Controller;
4 import se.kth.ict.oodbook.design.casestudy.dbhandler.CarDTO;
5
6 /**
7 * This program has no view, instead, this class is a placeholder
8 * for the entire view.
9 */
10 public class View {
11 public View(Controller contr) {
12 }
13 }

Listing C.65 Java code implementing the View class in figure 5.28

C.18 Figure 5.29

1 package se.kth.ict.oodbook.design.casestudy.view;
2
3 import se.kth.ict.oodbook.design.casestudy.controller.Controller;
4 import se.kth.ict.oodbook.design.casestudy.dbhandler.CarDTO;
5
6 /**
7 * This program has no view, instead, this class is a placeholder
8 * for the entire view.
9 */
10 public class View {
11 private Controller contr;
12
13 /**
14 * Creates a new instance.
15 *
16 * @param contr The controller that is used for all operations.
17 */
18 public View(Controller contr) {
19 }
20 }

262
 Appendix C Implementations of UML Diagrams

Listing C.66 Java code implementing the View class in figure 5.29

1 package se.kth.ict.oodbook.design.casestudy.controller;
2
3 import se.kth.ict.oodbook.design.casestudy.dbhandler.CarDTO;
4 import se.kth.ict.oodbook.design.casestudy.dbhandler.CarRegistry;
5
6 /**
7 * This is the application’s only controller class. All calls to
8 * the model pass through here.
9 */
10 public class Controller {
11 private CarRegistry carRegistry;
12
13 /**
14 * Creates a new instance.
15 *
16 * @param carRegistry Used to access the car data store.
17 */
18 public Controller(CarRegistry carRegistry) {
19 }
20
21 /**
22 * Search for a car matching the specified search criteria.
23 *
24 * @param searchedCar This object contains the search criteria.
25 * Fields in the object that are set to
26 * <code>null</code> or
27 * <code>false</code> are ignored.
28 * @return The best match of the search criteria.
29 */
30 public CarDTO searchMatchingCar(CarDTO searchedCar) {
31 }
32
33 /**
34 * Registers a new customer. Only registered customers can
35 * rent cars.
36 *
37 * @param customer The customer that will be registered.
38 */
39 public void registerCustomer(CustomerDTO customer) {
40 }
41 }

263
 Appendix C Implementations of UML Diagrams

Listing C.67 Java code implementing the Controller class in figure 5.29

1 package se.kth.ict.oodbook.design.casestudy.startup;
2
3 import se.kth.ict.oodbook.design.casestudy.controller.Controller;
4 import se.kth.ict.oodbook.design.casestudy.dbhandler.CarRegistry;
5 import se.kth.ict.oodbook.design.casestudy.view.View;
6
7 /**
8 * Contains the <code>main</code> method. Performs all startup
9 * of the application.
10 */
11 public class Main {
12 public static void main(String[] args) {
13 }
14 }

Listing C.68 Java code implementing the Main class in figure 5.29

1 package se.kth.ict.oodbook.design.casestudy.dbhandler;
2
3 /**
4 * Contains all calls to the data store with cars that may be
5 * rented.
6 */
7 public class CarRegistry {
8 /**
9 * Creates a new instance.
10 */
11 public CarRegistry() {
12 }
13
14 /**
15 * Search for a car matching the specified search criteria.
16 *
17 * @param searchedCar This object contains the search criteria.
18 * Fields in the object that are set to
19 * <code>null</code> or
20 * <code>false</code> are ignored.
21 * @return The best match of the search criteria.
22 */
23 public CarDTO findCar(CarDTO searchedCar) {
24 }

264
 Appendix C Implementations of UML Diagrams

25 }

Listing C.69 Java code implementing the CarRegistry class in figure 5.29

1 package se.kth.ict.oodbook.design.casestudy.dbhandler;
2
3 /**
4 * Contains information about one particular car.
5 */
6 public class CarDTO {
7
8 private int price;
9 private String size;
10 private boolean AC;
11 private boolean fourWD;
12 private String color;
13 private String regNo;
14
15 /**
16 * Creates a new instance representing a particular car.
17 *
18 * @param price The price paid to rent the car.
19 * @param size The size of the car, e.g.,
20 * <code>medium hatchback</code>.
21 * @param AC <code>true</code> if the car has
22 * air condition.
23 * @param fourWD <code>true</code> if the car has four
24 * wheel drive.
25 * @param color The color of the car.
26 * @param regNo The car’s registration number.
27 */
28 public CarDTO(int price, String size, boolean AC,
29 boolean fourWD, String color, String regNo) {
30 }
31
32 /**
33 * Get the value of regNo
34 *
35 * @return the value of regNo
36 */
37 public String getRegNo() {
38 }
39
40 /**
41 * Get the value of color
42 *

265
 Appendix C Implementations of UML Diagrams

43 * @return the value of color

44 */
45 public String getColor() {
46 }
47
48 /**
49 * Get the value of fourWD
50 *
51 * @return the value of fourWD
52 */
53 public boolean isFourWD() {
54 }
55
56 /**
57 * Get the value of AC
58 *
59 * @return the value of AC
60 */
61 public boolean isAC() {
62 }
63
64 /**
65 * Get the value of size
66 *
67 * @return the value of size
68 */
69 public String getSize() {
70 }
71
72 /**
73 * Get the value of price
74 *
75 * @return the value of price
76 */
77 public int getPrice() {
78 }
79
80 }

Listing C.70 Java code implementing the CarDTO class in figure 5.29

C.19 Figure 5.30

1 package se.kth.ict.oodbook.design.casestudy.view;

266
 Appendix C Implementations of UML Diagrams

2
3 import se.kth.ict.oodbook.design.casestudy.controller.Controller;
4 import se.kth.ict.oodbook.design.casestudy.dbhandler.CarDTO;
5 import se.kth.ict.oodbook.design.casestudy.model.AddressDTO;
6 import se.kth.ict.oodbook.design.casestudy.model.CustomerDTO;
7 import se.kth.ict.oodbook.design.casestudy.model.DrivingLicenseDTO;
8
9 /**
10 * This program has no view, instead, this class is a placeholder
11 * for the entire view.
12 */
13 public class View {
14 private Controller contr;
15
16 // Somewhere in the code. Note that the arguments to the
17 // DTO constructors are not specified in the UML
18 // diagram.
19 AddressDTO address = new AddressDTO("Storgatan 2", "12345",
20 "Hemorten");
21 DrivingLicenseDTO drivingLicense = new DrivingLicenseDTO(
22 "982193721937213");
23 CustomerDTO customer = new CustomerDTO("Stina", address,
24 drivingLicense);
25 contr.registerCustomer(customer);
26 }

Listing C.71 Java code implementing the View class in figure 5.30

1 package se.kth.ict.oodbook.design.casestudy.model;
2
3 /**
4 * Represents a post address.
5 */
6 public final class AddressDTO {
7 private final String street;
8 private final String zip;
9 private final String city;
10
11 /**
12 * Creates a new instance.
13 *
14 * @param street Street name and number.
15 * @param zip Zip code
16 * @param city City (postort)
17 */
18 public AddressDTO(String street, String zip, String city) {

267
 Appendix C Implementations of UML Diagrams

19 this.street = street;
20 this.zip = zip;
21 this.city = city;
22 }
23
24 /**
25 * Get the value of city
26 *
27 * @return the value of city
28 */
29 public String getCity() {
30 return city;
31 }
32
33 /**
34 * Get the value of zip
35 *
36 * @return the value of zip
37 */
38 public String getZip() {
39 return zip;
40 }
41
42 /**
43 * Get the value of street
44 *
45 * @return the value of street
46 */
47 public String getStreet() {
48 return street;
49 }
50
51 }

Listing C.72 Java code implementing the AddressDTO class in figure 5.30

1 package se.kth.ict.oodbook.design.casestudy.model;
2
3 /**
4 * Represents a driving license
5 */
6 public class DrivingLicenseDTO {
7 private final String licenseNo;
8
9 /**
10 * Creates a new instance.

268
 Appendix C Implementations of UML Diagrams

11 *
12 * @param licenseNo The driving license number.
13 */
14 public DrivingLicenseDTO(String licenseNo) {
15 this.licenseNo = licenseNo;
16 }
17
18 /**
19 * Get the value of licenseNo
20 *
21 * @return the value of licenseNo
22 */
23 public String getLicenseNo() {
24 return licenseNo;
25 }
26
27 }

Listing C.73 Java code implementing the DrivingLicenseDTO class in

figure 5.30

1 package se.kth.ict.oodbook.design.casestudy.model;
2
3 /**
4 * Represents a customer of the car rental company.
5 */
6 public class CustomerDTO {
7 private final String name;
8 private final AddressDTO address;
9 private final DrivingLicenseDTO drivingLicense;
10
11 /**
12 * Creates a new instance.
13 *
14 * @param name The customer’s name.
15 * @param address The customer’s address.
16 * @param drivingLicense The customer’s driving license.
17 */
18 public CustomerDTO(String name, AddressDTO address,
19 DrivingLicenseDTO drivingLicense) {
20 this.name = name;
21 this.address = address;
22 this.drivingLicense = drivingLicense;
23 }
24
25 /**

269
 Appendix C Implementations of UML Diagrams

26 * Get the value of drivingLicense

27 *
28 * @return the value of drivingLicense
29 */
30 public DrivingLicenseDTO getDrivingLicense() {
31 return drivingLicense;
32 }
33
34 /**
35 * Get the value of address
36 *
37 * @return the value of address
38 */
39 public AddressDTO getAddress() {
40 return address;
41 }
42
43 /**
44 * Get the value of name
45 *
46 * @return the value of name
47 */
48 public String getName() {
49 return name;
50 }
51
52 }

Listing C.74 Java code implementing the CustomerDTO class in figure 5.30

1 package se.kth.ict.oodbook.design.casestudy.controller;
2
3 import se.kth.ict.oodbook.design.casestudy.dbhandler.CarDTO;
4 import se.kth.ict.oodbook.design.casestudy.dbhandler.CarRegistry;
5 import se.kth.ict.oodbook.design.casestudy.model.CustomerDTO;
6 import se.kth.ict.oodbook.design.casestudy.model.Rental;
7
8 /**
9 * This is the application’s only controller class. All calls to
10 * the model pass through here.
11 */
12 public class Controller {
13 /**
14 * Registers a new customer. Only registered customers can
15 * rent cars.
16 *

270
 Appendix C Implementations of UML Diagrams

17 * @param customer The customer that will be registered.

18 */
19 public void registerCustomer(CustomerDTO customer) {
20 }
21 }

Listing C.75 Java code implementing the Controller class in figure 5.30

C.20 Figure 5.31

1 package se.kth.ict.oodbook.design.casestudy.view;
2
3 import se.kth.ict.oodbook.design.casestudy.controller.Controller;
4 import se.kth.ict.oodbook.design.casestudy.dbhandler.CarDTO;
5 import se.kth.ict.oodbook.design.casestudy.model.AddressDTO;
6 import se.kth.ict.oodbook.design.casestudy.model.CustomerDTO;
7 import se.kth.ict.oodbook.design.casestudy.model.DrivingLicenseDTO;
8
9 /**
10 * This program has no view, instead, this class is a placeholder
11 * for the entire view.
12 */
13 public class View {
14 private Controller contr;
15
16 // Somewhere in the code. Note that the arguments to the
17 // DTO constructors are not specified in the UML
18 // diagram.
19 AddressDTO address = new AddressDTO("Storgatan 2", "12345",
20 "Hemorten");
21 DrivingLicenseDTO drivingLicense = new DrivingLicenseDTO(
22 "982193721937213");
23 CustomerDTO customer = new CustomerDTO("Stina", address,
24 drivingLicense);
25 contr.registerCustomer(customer);
26 }

Listing C.76 Java code implementing the View class in figure 5.31

1 package se.kth.ict.oodbook.design.casestudy.model;
2
3 /**
4 * Represents a post address.

271
 Appendix C Implementations of UML Diagrams

5 */
6 public final class AddressDTO {
7 private final String street;
8 private final String zip;
9 private final String city;
10
11 /**
12 * Creates a new instance.
13 *
14 * @param street Street name and number.
15 * @param zip Zip code
16 * @param city City (postort)
17 */
18 public AddressDTO(String street, String zip, String city) {
19 this.street = street;
20 this.zip = zip;
21 this.city = city;
22 }
23
24 /**
25 * Get the value of city
26 *
27 * @return the value of city
28 */
29 public String getCity() {
30 return city;
31 }
32
33 /**
34 * Get the value of zip
35 *
36 * @return the value of zip
37 */
38 public String getZip() {
39 return zip;
40 }
41
42 /**
43 * Get the value of street
44 *
45 * @return the value of street
46 */
47 public String getStreet() {
48 return street;
49 }
50

272
 Appendix C Implementations of UML Diagrams

51 }

Listing C.77 Java code implementing the AddressDTO class in figure 5.31

1 package se.kth.ict.oodbook.design.casestudy.model;
2
3 /**
4 * Represents a driving license
5 */
6 public class DrivingLicenseDTO {
7 private final String licenseNo;
8
9 /**
10 * Creates a new instance.
11 *
12 * @param licenseNo The driving license number.
13 */
14 public DrivingLicenseDTO(String licenseNo) {
15 this.licenseNo = licenseNo;
16 }
17
18 /**
19 * Get the value of licenseNo
20 *
21 * @return the value of licenseNo
22 */
23 public String getLicenseNo() {
24 return licenseNo;
25 }
26
27 }

Listing C.78 Java code implementing the DrivingLicenseDTO class in

figure 5.31

1 package se.kth.ict.oodbook.design.casestudy.model;
2
3 /**
4 * Represents a customer of the car rental company.
5 */
6 public class CustomerDTO {
7 private final String name;
8 private final AddressDTO address;
9 private final DrivingLicenseDTO drivingLicense;
10
11 /**

273
 Appendix C Implementations of UML Diagrams

12 * Creates a new instance.

13 *
14 * @param name The customer’s name.
15 * @param address The customer’s address.
16 * @param drivingLicense The customer’s driving license.
17 */
18 public CustomerDTO(String name, AddressDTO address,
19 DrivingLicenseDTO drivingLicense) {
20 this.name = name;
21 this.address = address;
22 this.drivingLicense = drivingLicense;
23 }
24
25 /**
26 * Get the value of drivingLicense
27 *
28 * @return the value of drivingLicense
29 */
30 public DrivingLicenseDTO getDrivingLicense() {
31 return drivingLicense;
32 }
33
34 /**
35 * Get the value of address
36 *
37 * @return the value of address
38 */
39 public AddressDTO getAddress() {
40 return address;
41 }
42
43 /**
44 * Get the value of name
45 *
46 * @return the value of name
47 */
48 public String getName() {
49 return name;
50 }
51
52 }

Listing C.79 Java code implementing the CustomerDTO class in figure 5.31

1 package se.kth.ict.oodbook.design.casestudy.controller;
2

274
 Appendix C Implementations of UML Diagrams

3 import se.kth.ict.oodbook.design.casestudy.dbhandler.CarDTO;
4 import se.kth.ict.oodbook.design.casestudy.dbhandler.CarRegistry;
5 import se.kth.ict.oodbook.design.casestudy.model.CustomerDTO;
6 import se.kth.ict.oodbook.design.casestudy.model.Rental;
7
8 /**
9 * This is the application’s only controller class. All calls to
10 * the model pass through here.
11 */
12 public class Controller {
13 /**
14 * Registers a new customer. Only registered customers can
15 * rent cars.
16 *
17 * @param customer The customer that will be registered.
18 */
19 public void registerCustomer(CustomerDTO customer) {
20 rental = new Rental(customer);
21 }
22 }

Listing C.80 Java code implementing the Controller class in figure 5.31

1 package se.kth.ict.oodbook.design.casestudy.model;
2
3 /**
4 * Represents one particular rental transaction, where one
5 * particular car is rented by one particular customer.
6 */
7 public class Rental {
8 private CustomerDTO customer;
9
10 /**
11 * Creates a new instance, representing a rental made by the
12 * specified customer.
13 *
14 * @param customer The renting customer.
15 */
16 public Rental(CustomerDTO customer) {
17 this.customer = customer;
18 }
19 }

Listing C.81 Java code implementing the Rental class in figure 5.31

275
 Appendix C Implementations of UML Diagrams

C.21 Figure 5.32

1 package se.kth.ict.oodbook.design.casestudy.view;
2
3 import se.kth.ict.oodbook.design.casestudy.controller.Controller;
4 import se.kth.ict.oodbook.design.casestudy.dbhandler.CarDTO;
5
6 /**
7 * This program has no view, instead, this class is a placeholder
8 * for the entire view.
9 */
10 public class View {
11 private Controller contr;
12
13 /**
14 * Creates a new instance.
15 *
16 * @param contr The controller that is used for all operations.
17 */
18 public View(Controller contr) {
19 }
20 }

Listing C.82 Java code implementing the View class in figure 5.32

276
 Appendix C Implementations of UML Diagrams

1 package se.kth.ict.oodbook.design.casestudy.controller;
2
3 import se.kth.ict.oodbook.design.casestudy.dbhandler.CarDTO;
4 import se.kth.ict.oodbook.design.casestudy.dbhandler.CarRegistry;
5
6 /**
7 * This is the application’s only controller class. All calls to
8 * the model pass through here.
9 */
10 public class Controller {
11 private CarRegistry carRegistry;
12
13 /**
14 * Creates a new instance.
15 *
16 * @param carRegistry Used to access the car data store.
17 */
18 public Controller(CarRegistry carRegistry) {
19 }
20
21 /**
22 * Search for a car matching the specified search criteria.
23 *
24 * @param searchedCar This object contains the search criteria.
25 * Fields in the object that are set to
26 * <code>null</code> or
27 * <code>false</code> are ignored.
28 * @return The best match of the search criteria.
29 */
30 public CarDTO searchMatchingCar(CarDTO searchedCar) {
31 }
32
33 /**
34 * Registers a new customer. Only registered customers can
35 * rent cars.
36 *
37 * @param customer The customer that will be registered.
38 */
39 public void registerCustomer(CustomerDTO customer) {
40 }
41 }

Listing C.83 Java code implementing the Controller class in figure 5.32

1 package se.kth.ict.oodbook.design.casestudy.startup;

277
 Appendix C Implementations of UML Diagrams

2
3 import se.kth.ict.oodbook.design.casestudy.controller.Controller;
4 import se.kth.ict.oodbook.design.casestudy.dbhandler.CarRegistry;
5 import se.kth.ict.oodbook.design.casestudy.view.View;
6
7 /**
8 * Contains the <code>main</code> method. Performs all startup
9 * of the application.
10 */
11 public class Main {
12 public static void main(String[] args) {
13 }
14 }

Listing C.84 Java code implementing the Main class in figure 5.32

1 package se.kth.ict.oodbook.design.casestudy.dbhandler;
2
3 /**
4 * Contains all calls to the data store with cars that may be
5 * rented.
6 */
7 public class CarRegistry {
8 /**
9 * Creates a new instance.
10 */
11 public CarRegistry() {
12 }
13
14 /**
15 * Search for a car matching the specified search criteria.
16 *
17 * @param searchedCar This object contains the search
18 * criteria. Fields in the object that
19 * are set to <code>null</code> or
20 * <code>false</code> are ignored.
21 * @return The best match of the search criteria.
22 */
23 public CarDTO findCar(CarDTO searchedCar) {
24 }
25 }

Listing C.85 Java code implementing the CarRegistry class in figure 5.32

1 package se.kth.ict.oodbook.design.casestudy.dbhandler;

278
 Appendix C Implementations of UML Diagrams

2
3 /**
4 * Contains information about one particular car.
5 */
6 public class CarDTO {
7
8 private int price;
9 private String size;
10 private boolean AC;
11 private boolean fourWD;
12 private String color;
13 private String regNo;
14
15 /**
16 * Creates a new instance representing a particular car.
17 *
18 * @param price The price paid to rent the car.
19 * @param size The size of the car, e.g.,
20 * <code>medium hatchback</code>.
21 * @param AC <code>true</code> if the car has
22 * air condition.
23 * @param fourWD <code>true</code> if the car has four
24 * wheel drive.
25 * @param color The color of the car.
26 * @param regNo The car’s registration number.
27 */
28 public CarDTO(int price, String size, boolean AC,
29 boolean fourWD, String color, String regNo) {
30 }
31
32 /**
33 * Get the value of regNo
34 *
35 * @return the value of regNo
36 */
37 public String getRegNo() {
38 }
39
40 /**
41 * Get the value of color
42 *
43 * @return the value of color
44 */
45 public String getColor() {
46 }
47

279
 Appendix C Implementations of UML Diagrams

48 /**
49 * Get the value of fourWD
50 *
51 * @return the value of fourWD
52 */
53 public boolean isFourWD() {
54 }
55
56 /**
57 * Get the value of AC
58 *
59 * @return the value of AC
60 */
61 public boolean isAC() {
62 }
63
64 /**
65 * Get the value of size
66 *
67 * @return the value of size
68 */
69 public String getSize() {
70 }
71
72 /**
73 * Get the value of price
74 *
75 * @return the value of price
76 */
77 public int getPrice() {
78 }
79
80 }

Listing C.86 Java code implementing the CarDTO class in figure 5.32

1 package se.kth.ict.oodbook.design.casestudy.model;
2
3 /**
4 * Represents one particular rental transaction, where one
5 * particular car is rented by one particular customer.
6 */
7 public class Rental {
8 private CustomerDTO customer;
9
10 /**

280
 Appendix C Implementations of UML Diagrams

11 * Creates a new instance, representing a rental made by

12 * the specified customer.
13 *
14 * @param customer The renting customer.
15 */
16 public Rental(CustomerDTO customer) {
17 }
18 }

Listing C.87 Java code implementing the Rental class in figure 5.32

1 package se.kth.ict.oodbook.design.casestudy.model;
2
3 /**
4 * Represents a driving license
5 */
6 public class DrivingLicenseDTO {
7 private final String licenseNo;
8
9 /**
10 * Creates a new instance.
11 *
12 * @param licenseNo The driving license number.
13 */
14 public DrivingLicenseDTO(String licenseNo) {
15 }
16
17 /**
18 * Get the value of licenseNo
19 *
20 * @return the value of licenseNo
21 */
22 public String getLicenseNo() {
23 }
24
25 }

Listing C.88 Java code implementing the DrivingLicenseDTO class in

figure 5.32

1 package se.kth.ict.oodbook.design.casestudy.model;
2
3 /**
4 * Represents a customer of the car rental company.
5 */
6 public class CustomerDTO {

281
 Appendix C Implementations of UML Diagrams

7 private final String name;

8 private final AddressDTO address;
9 private final DrivingLicenseDTO drivingLicense;
10
11 /**
12 * Creates a new instance.
13 *
14 * @param name The customer’s name.
15 * @param address The customer’s address.
16 * @param drivingLicense The customer’s driving license.
17 */
18 public CustomerDTO(String name, AddressDTO address,
19 DrivingLicenseDTO drivingLicense) {
20 }
21
22 /**
23 * Get the value of drivingLicense
24 *
25 * @return the value of drivingLicense
26 */
27 public DrivingLicenseDTO getDrivingLicense() {
28 }
29
30 /**
31 * Get the value of address
32 *
33 * @return the value of address
34 */
35 public AddressDTO getAddress() {
36 }
37
38 /**
39 * Get the value of name
40 *
41 * @return the value of name
42 */
43 public String getName() {
44 }
45
46 }

Listing C.89 Java code implementing the CustomerDTO class in figure 5.32

1 package se.kth.ict.oodbook.design.casestudy.model;
2
3 /**

282
 Appendix C Implementations of UML Diagrams

4 * Represents a post address.

5 */
6 public final class AddressDTO {
7 private final String street;
8 private final String zip;
9 private final String city;
10
11 /**
12 * Creates a new instance.
13 *
14 * @param street Street name and number.
15 * @param zip Zip code
16 * @param city City (postort)
17 */
18 public AddressDTO(String street, String zip, String city) {
19 }
20
21 /**
22 * Get the value of city
23 *
24 * @return the value of city
25 */
26 public String getCity() {
27 }
28
29 /**
30 * Get the value of zip
31 *
32 * @return the value of zip
33 */
34 public String getZip() {
35 }
36
37 /**
38 * Get the value of street
39 *
40 * @return the value of street
41 */
42 public String getStreet() {
43 }
44
45 }

Listing C.90 Java code implementing the AddressDTO class in figure 5.32

283
 Appendix C Implementations of UML Diagrams

C.22 Figure 5.33

1 package se.kth.ict.oodbook.design.casestudy.view;
2
3 import se.kth.ict.oodbook.design.casestudy.controller.Controller;
4 import se.kth.ict.oodbook.design.casestudy.dbhandler.CarDTO;
5 import se.kth.ict.oodbook.design.casestudy.model.AddressDTO;
6 import se.kth.ict.oodbook.design.casestudy.model.CustomerDTO;
7 import se.kth.ict.oodbook.design.casestudy.model.DrivingLicenseDTO;
8
9 /**
10 * This program has no view, instead, this class is a placeholder
11 * for the entire view.
12 */
13 public class View {
14 private Controller contr;
15
16 //Somewhere in the code.
17 contr.bookCar(foundCar);
18 }

Listing C.91 Java code implementing the View class in figure 5.33

1 package se.kth.ict.oodbook.design.casestudy.controller;
2
3 import se.kth.ict.oodbook.design.casestudy.dbhandler.CarDTO;
4 import se.kth.ict.oodbook.design.casestudy.dbhandler.CarRegistry;
5 import se.kth.ict.oodbook.design.casestudy.dbhandler.RentalRegistry;
6 import se.kth.ict.oodbook.design.casestudy.model.CustomerDTO;
7 import se.kth.ict.oodbook.design.casestudy.model.Rental;
8
9 /**
10 * This is the application’s only controller class. All calls to the
11 * model pass through here.
12 */
13 public class Controller {
14 private RentalRegistry rentalRegistry;
15 private Rental rental;
16
17 /**
18 * Books the specified car. After calling this method, the car
19 * can not be booked by any other customer. This method also
20 * permanently saves information about the current rental.
21 *
22 * @param car The car that will be booked.

284
 Appendix C Implementations of UML Diagrams

23 */
24 public void bookCar(CarDTO car) {
25 rental.setRentedCar(car);
26 rentalRegistry.saveRental(rental);
27 }
28 }

Listing C.92 Java code implementing the Controller class in figure 5.33

1 package se.kth.ict.oodbook.design.casestudy.model;
2
3 import se.kth.ict.oodbook.design.casestudy.dbhandler.CarDTO;
4 import se.kth.ict.oodbook.design.casestudy.dbhandler.CarRegistry;
5
6 /**
7 * Represents one particular rental transaction, where one
8 * particular car is rented by one particular customer.
9 */
10 public class Rental {
11 private CarDTO rentedCar;
12 private CarRegistry carRegistry;
13 /**
14 * Specifies the car that was rented.
15 *
16 * @param rentedCar The car that was rented.
17 */
18 public void setRentedCar(CarDTO rentedCar) {
19 this.rentedCar = rentedCar;
20 carRegistry.bookCar(rentedCar);
21 }
22 }

Listing C.93 Java code implementing the Rental class in figure 5.33

1 package se.kth.ict.oodbook.design.casestudy.dbhandler;
2
3 import se.kth.ict.oodbook.design.casestudy.model.Rental;
4
5 /**
6 * Contains all calls to the data store with performed rentals.
7 */
8 public class RentalRegistry {
9 /**
10 * Saves the specified rental permanently.
11 *

285
 Appendix C Implementations of UML Diagrams

12 * @param rental The rental that will be saved.

13 */
14 public void saveRental(Rental rental) {
15 }
16 }

Listing C.94 Java code implementing the RentalRegistry class in figure 5.33

1 package se.kth.ict.oodbook.design.casestudy.dbhandler;
2
3 /**
4 * Contains all calls to the data store with cars that may
5 * be rented.
6 */
7 public class CarRegistry {
8 /**
9 * Books the specified car. After calling this method,
10 * the car can not be booked by any other customer.
11 *
12 * @param car The car that will be booked.
13 */
14 public void bookCar(CarDTO car) {
15 }
16 }

Listing C.95 Java code implementing the CarRegistry class in figure 5.33

C.23 Figure 5.34

1 package se.kth.ict.oodbook.design.casestudy.startup;
2
3 import se.kth.ict.oodbook.design.casestudy.controller.Controller;
4 import se.kth.ict.oodbook.design.casestudy.dbhandler.CarRegistry;
5 import se.kth.ict.oodbook.design.casestudy.view.View;
6
7 /**
8 * Contains the <code>main</code> method. Performs all startup of
9 * the application.
10 */
11 public class Main {
12 public static void main(String[] args) {
13 CarRegistry carRegistry = new CarRegistry();
14 RentalRegistry rentalRegistry = new RentalRegistry();

286
 Appendix C Implementations of UML Diagrams

15 Controller contr = new Controller(carRegistry,

16 rentalRegistry);
17 new View(contr);
18 }
19 }

Listing C.96 Java code implementing the Main class in figure 5.34

1 package se.kth.ict.oodbook.design.casestudy.dbhandler;
2
3 /**
4 * Contains all calls to the data store with cars that may be
5 * rented.
6 */
7 public class CarRegistry {
8 }

Listing C.97 Java code implementing the CarRegistry class in figure 5.34

1 package se.kth.ict.oodbook.design.casestudy.dbhandler;
2
3 import se.kth.ict.oodbook.design.casestudy.model.Rental;
4
5 /**
6 * Contains all calls to the data store with performed rentals.
7 */
8 public class RentalRegistry {
9 }

Listing C.98 Java code implementing the RentalRegistry class in figure 5.34

1 package se.kth.ict.oodbook.design.casestudy.controller;
2
3 import se.kth.ict.oodbook.design.casestudy.dbhandler.CarDTO;
4 import se.kth.ict.oodbook.design.casestudy.dbhandler.CarRegistry;
5
6 /**
7 * This is the application’s only controller class. All calls to
8 * the model pass through here.
9 */
10 public class Controller {
11 /**
12 * Creates a new instance.
13 *

287
 Appendix C Implementations of UML Diagrams

14 * @param carRegistry Used to access the car data store.

15 * @param rentalRegistry Used to access the rental data store.
16 */
17 public Controller(CarRegistry carRegistry,
18 RentalRegistry rentalRegistry) {
19 }
20 }

Listing C.99 Java code implementing the Controller class in figure 5.34

1 package se.kth.ict.oodbook.design.casestudy.view;
2
3 import se.kth.ict.oodbook.design.casestudy.controller.Controller;
4 import se.kth.ict.oodbook.design.casestudy.dbhandler.CarDTO;
5
6 /**
7 * This program has no view, instead, this class is a placeholder
8 * for the entire view.
9 */
10 public class View {
11 /**
12 * Creates a new instance.
13 *
14 * @param contr The controller that is used for all
15 * operations.
16 */
17 public View(Controller contr) {
18 }
19 }

Listing C.100 Java code implementing the View class in figure 5.34

C.24 Figure 5.35

1 package se.kth.ict.oodbook.design.casestudy.startup;
2
3 import se.kth.ict.oodbook.design.casestudy.controller.Controller;
4 import se.kth.ict.oodbook.design.casestudy.dbhandler.CarRegistry;
5 import se.kth.ict.oodbook.design.casestudy.view.View;
6
7 /**
8 * Contains the <code>main</code> method. Performs all startup of
9 * the application.

288
 Appendix C Implementations of UML Diagrams

10 */
11 public class Main {
12 public static void main(String[] args) {
13 RegistryCreator creator = new RegistryCreator();
14 Controller contr = new Controller(creator);
15 new View(contr);
16 }
17 }

Listing C.101 Java code implementing the Main class in figure 5.35

1 package se.kth.ict.oodbook.design.casestudy.dbhandler;
2
3 /**
4 * Contains all calls to the data store with cars that may be
5 * rented.
6 */
7 public class CarRegistry {
8 }

Listing C.102 Java code implementing the CarRegistry class in figure 5.35

1 package se.kth.ict.oodbook.design.casestudy.dbhandler;
2
3 import se.kth.ict.oodbook.design.casestudy.model.Rental;
4
5 /**
6 * Contains all calls to the data store with performed rentals.
7 */
8 public class RentalRegistry {
9 }

Listing C.103 Java code implementing the RentalRegistry class in figure 5.35

1 package se.kth.ict.oodbook.design.casestudy.dbhandler;
2
3 /**
4 * This class is responsible for instantiating all registries.
5 */
6 public class RegistryCreator {
7 private CarRegistry carRegistry = new CarRegistry();
8 private RentalRegistry rentalRegistry = new RentalRegistry();
9
10 /**

289
 Appendix C Implementations of UML Diagrams

11 * Get the value of rentalRegistry

12 *
13 * @return the value of rentalRegistry
14 */
15 public RentalRegistry getRentalRegistry() {
16 return rentalRegistry;
17 }
18
19 /**
20 * Get the value of carRegistry
21 *
22 * @return the value of carRegistry
23 */
24 public CarRegistry getCarRegistry() {
25 return carRegistry;
26 }
27 }

Listing C.104 Java code implementing the RegistryCreator class in figure 5.35

1 package se.kth.ict.oodbook.design.casestudy.controller;
2
3 import se.kth.ict.oodbook.design.casestudy.dbhandler.CarDTO;
4 import se.kth.ict.oodbook.design.casestudy.dbhandler.CarRegistry;
5
6 /**
7 * This is the application’s only controller class. All calls to
8 * the model pass through here.
9 */
10 public class Controller {
11 /**
12 * Creates a new instance.
13 *
14 * @param regCreator Used to get all classes that handle
15 * database calls.
16 */
17 public Controller(RegistryCreator regCreator) {
18 }
19 }

Listing C.105 Java code implementing the Controller class in figure 5.35

1 package se.kth.ict.oodbook.design.casestudy.view;
2
3 import se.kth.ict.oodbook.design.casestudy.controller.Controller;

290
 Appendix C Implementations of UML Diagrams

4 import se.kth.ict.oodbook.design.casestudy.dbhandler.CarDTO;
5
6 /**
7 * This program has no view, instead, this class is a placeholder
8 * for the entire view.
9 */
10 public class View {
11 /**
12 * Creates a new instance.
13 *
14 * @param contr The controller that is used for all
15 * operations.
16 */
17 public View(Controller contr) {
18 }
19 }

Listing C.106 Java code implementing the View class in figure 5.35

C.25 Figure 5.36

1 package se.kth.ict.oodbook.design.casestudy.view;
2
3 import se.kth.ict.oodbook.design.casestudy.controller.Controller;
4
5 /**
6 * This program has no view, instead, this class is a placeholder
7 * for the entire view.
8 */
9 public class View {
10 private Controller contr;
11
12 /**
13 * Creates a new instance.
14 *
15 * @param contr The controller that is used for all
16 * operations.
17 */
18 public View(Controller contr) {
19 }
20 }

Listing C.107 Java code implementing the View class in figure 5.36

291
 Appendix C Implementations of UML Diagrams

1 package se.kth.ict.oodbook.design.casestudy.controller;
2
3 import se.kth.ict.oodbook.design.casestudy.dbhandler.CarDTO;
4 import se.kth.ict.oodbook.design.casestudy.dbhandler.CarRegistry;
5 import se.kth.ict.oodbook.design.casestudy.dbhandler.RegistryCreator;
6 import se.kth.ict.oodbook.design.casestudy.dbhandler.RentalRegistry;
7 import se.kth.ict.oodbook.design.casestudy.model.CustomerDTO;
8 import se.kth.ict.oodbook.design.casestudy.model.Rental;
9
10 /**
11 * This is the application’s only controller class. All calls to
12 * the model pass through here.
13 */
14 public class Controller {
15 private CarRegistry carRegistry;
16 private RentalRegistry rentalRegistry;
17 private Rental rental;
18
19 /**
20 * Creates a new instance.
21 *
22 * @param regCreator Used to get all classes that handle
23 * database calls.
24 */
25 public Controller(RegistryCreator regCreator) {
26 }
27
28 /**
29 * Search for a car matching the specified search criteria.
30 *
31 * @param searchedCar This object contains the search criteria.
32 * Fields in the object that are set to
33 * <code>null</code> or
34 * <code>false</code> are ignored.
35 * @return The best match of the search criteria.
36 */
37 public CarDTO searchMatchingCar(CarDTO searchedCar) {
38 }
39
40 /**
41 * Registers a new customer. Only registered customers can
42 * rent cars.
43 *
44 * @param customer The customer that will be registered.
45 */
46 public void registerCustomer(CustomerDTO customer) {

292
 Appendix C Implementations of UML Diagrams

47 }
48
49 /**
50 * Books the specified car. After calling this method, the car
51 * can not be booked by any other customer. This method also
52 * permanently saves information about the current rental.
53 *
54 * @param car The car that will be booked.
55 */
56 public void bookCar(CarDTO car) {
57 }
58 }

Listing C.108 Java code implementing the Controller class in figure 5.36

1 package se.kth.ict.oodbook.design.casestudy.startup;
2
3 import se.kth.ict.oodbook.design.casestudy.controller.Controller;
4 import se.kth.ict.oodbook.design.casestudy.dbhandler.RegistryCreator;
5 import se.kth.ict.oodbook.design.casestudy.view.View;
6
7 /**
8 * Contains the <code>main</code> method. Performs all startup of
9 * the application.
10 */
11 public class Main {
12 /**
13 * Starts the application.
14 *
15 * @param args The application does not take any command line
16 * parameters.
17 */
18 public static void main(String[] args) {
19 }
20 }

Listing C.109 Java code implementing the Main class in figure 5.36

1 package se.kth.ict.oodbook.design.casestudy.model;
2
3 import se.kth.ict.oodbook.design.casestudy.dbhandler.CarDTO;
4 import se.kth.ict.oodbook.design.casestudy.dbhandler.CarRegistry;
5
6 /**
7 * Represents one particular rental transaction, where one

293
 Appendix C Implementations of UML Diagrams

8 * particular car is rented by one particular customer.

9 */
10 public class Rental {
11 private CarRegistry carRegistry;
12
13 /**
14 * Creates a new instance, representing a rental made by the
15 * specified customer.
16 *
17 * @param customer The renting customer.
18 * @param carRegistry The data store with information about
19 * available cars.
20 */
21 public Rental(CustomerDTO customer, CarRegistry carRegistry) {
22 }
23
24 /**
25 * Specifies the car that was rented.
26 *
27 * @param rentedCar The car that was rented.
28 */
29 public void setRentedCar(CarDTO rentedCar) {
30 }
31 }

Listing C.110 Java code implementing the Rental class in figure 5.36

1 package se.kth.ict.oodbook.design.casestudy.dbhandler;
2
3 /**
4 * This class is responsible for instantiating all registries.
5 */
6 public class RegistryCreator {
7 /**
8 * Get the value of rentalRegistry
9 *
10 * @return the value of rentalRegistry
11 */
12 public RentalRegistry getRentalRegistry() {
13 }
14
15 /**
16 * Get the value of carRegistry
17 *
18 * @return the value of carRegistry
19 */

294
 Appendix C Implementations of UML Diagrams

20 public CarRegistry getCarRegistry() {

21 }
22 }

Listing C.111 Java code implementing the RegistryCreator class in figure 5.36

1 package se.kth.ict.oodbook.design.casestudy.dbhandler;
2
3 /**
4 * Contains all calls to the data store with cars that may be
5 * rented.
6 */
7 public class CarRegistry {
8 CarRegistry() {
9 }
10
11 /**
12 * Search for a car matching the specified search criteria.
13 *
14 * @param searchedCar This object contains the search criteria.
15 * Fields in the object that are set to
16 * <code>null</code> or <code>false</code>
17 * are ignored.
18 * @return The best match of the search criteria.
19 */
20 public CarDTO findCar(CarDTO searchedCar) {
21 }
22
23 /**
24 * Books the specified car. After calling this method, the car
25 * can not be booked by any other customer.
26 *
27 * @param car The car that will be booked.
28 */
29 public void bookCar(CarDTO car) {
30 }
31 }

Listing C.112 Java code implementing the CarRegistry class in figure 5.36

1 package se.kth.ict.oodbook.design.casestudy.dbhandler;
2
3 import se.kth.ict.oodbook.design.casestudy.model.Rental;
4
5 /**

295
 Appendix C Implementations of UML Diagrams

6 * Contains all calls to the data store with performed rentals.

7 */
8 public class RentalRegistry {
9 RentalRegistry() {
10 }
11
12 /**
13 * Saves the specified rental permanently.
14 *
15 * @param rental The rental that will be saved.
16 */
17 public void saveRental(Rental rental) {
18 }
19 }

Listing C.113 Java code implementing the RentalRegistry class in figure 5.36

C.26 Figure 5.37

1 package se.kth.ict.oodbook.design.casestudy.view;
2
3 import se.kth.ict.oodbook.design.casestudy.controller.Controller;
4 import se.kth.ict.oodbook.design.casestudy.model.Amount;
5
6 /**
7 * This program has no view, instead, this class is a placeholder
8 * for the entire view.
9 */
10 public class View {
11 private Controller contr;
12
13 // Somewhere in the code. The used amount (100) is not
14 // specified in the diagram.
15 Amount paidAmount = new Amount(100);
16 contr.pay(paidAmount);
17 }

Listing C.114 Java code implementing the View class in figure 5.37

1 package se.kth.ict.oodbook.design.casestudy.controller;
2
3 import se.kth.ict.oodbook.design.casestudy.model.Amount;
4 import se.kth.ict.oodbook.design.casestudy.model.Rental;

296
 Appendix C Implementations of UML Diagrams

5
6 /**
7 * This is the application’s only controller class. All
8 * calls to the model pass through here.
9 */
10 public class Controller {
11 /**
12 * Handles rental payment. Updates the balance of
13 * the cash register where the payment was
14 * performed. Calculates change. Prints the receipt.
15 *
16 * @param amount The paid amount.
17 */
18 public void pay(Amount paidAmt) {
19 }
20 }

Listing C.115 Java code implementing the Controller class in figure 5.37

1 package se.kth.ict.oodbook.design.casestudy.model;
2
3 /**
4 * Represents an amount of money
5 */
6 public final class Amount {
7 private final int amount;
8
9 public Amount(int amount) {
10 this.amount = amount;
11 }
12 }

Listing C.116 Java code implementing the Amount class in figure 5.37

C.27 Figure 5.38

1 package se.kth.ict.oodbook.design.casestudy.view;
2
3 import se.kth.ict.oodbook.design.casestudy.controller.Controller;
4 import se.kth.ict.oodbook.design.casestudy.model.Amount;
5
6 /**
7 * This program has no view, instead, this class is a placeholder

297
 Appendix C Implementations of UML Diagrams

8 * for the entire view.

9 */
10 public class View {
11 private Controller contr;
12
13 // Somewhere in the code. The used amount (100) is not
14 // specified in the diagram.
15 Amount paidAmount = new Amount(100);
16 contr.pay(paidAmount);
17 }

Listing C.117 Java code implementing the View class in figure 5.38

1 package se.kth.ict.oodbook.design.casestudy.controller;
2
3 import se.kth.ict.oodbook.design.casestudy.model.Amount;
4 import se.kth.ict.oodbook.design.casestudy.model.Rental;
5
6 /**
7 * This is the application’s only controller class. All
8 * calls to the model pass through here.
9 */
10 public class Controller {
11 /**
12 * Handles rental payment. Updates the balance of
13 * the cash register where the payment was
14 * performed. Calculates change. Prints the receipt.
15 *
16 * @param amount The paid amount.
17 */
18 public void pay(Amount paidAmt) {
19 CashPayment payment = new CashPayment(paidAmt);
20 rental.pay(payment);
21 cashRegister.addPayment(payment);
22 }
23 }

Listing C.118 Java code implementing the Controller class in figure 5.38

1 package se.kth.ict.oodbook.design.casestudy.model;
2
3 /**
4 * Represents a cash register. There shall be one instance of
5 * this class for each register.
6 */

298
 Appendix C Implementations of UML Diagrams

7 public class CashRegister {

8 public void addPayment(CashPayment payment) {
9 }
10 }

Listing C.119 Java code implementing the CashRegister class in figure 5.38

1 package se.kth.ict.oodbook.design.casestudy.model;
2
3 /**
4 * Represents one specific payment for one specific rental. The
5 * rental is payed with cash.
6 */
7 public class CashPayment {
8 private Amount paidAmt;
9
10 /**
11 * Creates a new instance. The customer handed over the
12 * specified amount.
13 *
14 * @param paidAmt The amount of cash that was handed over
15 * by the customer.
16 */
17 public CashPayment(Amount paidAmt) {
18 this.paidAmt = paidAmt;
19 }
20
21 /**
22 * Calculates the total cost of the specified rental.
23 *
24 * @param paidRental The rental for which the customer is
25 * paying.
26 */
27 void calculateTotalCost(Rental paidRental) {
28 }
29 }

Listing C.120 Java code implementing the CashPayment class in figure 5.38

1 package se.kth.ict.oodbook.design.casestudy.model;
2
3 import se.kth.ict.oodbook.design.casestudy.dbhandler.CarDTO;
4 import se.kth.ict.oodbook.design.casestudy.dbhandler.CarRegistry;
5
6 /**

299
 Appendix C Implementations of UML Diagrams

7 * Represents one particular rental transaction, where one

8 * particular car is rented by one particular customer.
9 */
10 public class Rental {
11 /**
12 * This rental is paid using the specified payment.
13 *
14 * @param payment The payment used to pay this rental.
15 */
16 public void pay(CashPayment payment) {
17 payment.calculateTotalCost(this);
18 }
19 }

Listing C.121 Java code implementing the Rental class in figure 5.38

1 package se.kth.ict.oodbook.design.casestudy.model;
2
3 /**
4 * Represents an amount of money
5 */
6 public final class Amount {
7 private final int amount;
8
9 public Amount(int amount) {
10 this.amount = amount;
11 }
12 }

Listing C.122 Java code implementing the Amount class in figure 5.38

C.28 Figure 5.39

1 package se.kth.ict.oodbook.design.casestudy.view;
2
3 import se.kth.ict.oodbook.design.casestudy.controller.Controller;
4 import se.kth.ict.oodbook.design.casestudy.model.Amount;
5
6 /**
7 * This program has no view, instead, this class is a placeholder
8 * for the entire view.
9 */
10 public class View {

300
 Appendix C Implementations of UML Diagrams

11 private Controller contr;

12
13 // Somewhere in the code. The used amount (100) is not
14 // specified in the diagram.
15 Amount paidAmount = new Amount(100);
16 contr.pay(paidAmount);
17 }

Listing C.123 Java code implementing the View class in figure 5.39

1 package se.kth.ict.oodbook.design.casestudy.controller;
2
3 import se.kth.ict.oodbook.design.casestudy.model.Amount;
4 import se.kth.ict.oodbook.design.casestudy.integration.Printer;
5 import se.kth.ict.oodbook.design.casestudy.model.CashPayment;
6 import se.kth.ict.oodbook.design.casestudy.model.CashRegister;
7 import se.kth.ict.oodbook.design.casestudy.model.Rental;
8 import se.kth.ict.oodbook.design.casestudy.model.Receipt;
9
10 /**
11 * This is the application’s only controller class. All
12 * calls to the model pass through here.
13 */
14 public class Controller {
15 private Rental rental;
16 private CashRegister cashRegister;
17 private Printer printer;
18
19 /**
20 * Handles rental payment. Updates the balance of
21 * the cash register where the payment was
22 * performed. Calculates change. Prints the receipt.
23 *
24 * @param amount The paid amount.
25 */
26 public void pay(Amount amount) {
27 CashPayment payment = new CashPayment(paidAmt);
28 rental.pay(payment);
29 cashRegister.addPayment(payment);
30 Receipt receipt = rental.getReceipt();
31 printer.printReceipt(receipt);
32 }
33 }

Listing C.124 Java code implementing the Controller class in figure 5.39

301
 Appendix C Implementations of UML Diagrams

1 package se.kth.ict.oodbook.design.casestudy.model;
2
3 /**
4 * Represents a cash register. There shall be one instance of
5 * this class for each register.
6 */
7 public class CashRegister {
8 public void addPayment(CashPayment payment) {
9 }
10 }

Listing C.125 Java code implementing the CashRegister class in figure 5.39

1 package se.kth.ict.oodbook.design.casestudy.model;
2
3 /**
4 * Represents one specific payment for one specific rental. The
5 * rental is payed with cash.
6 */
7 public class CashPayment {
8 private Amount paidAmt;
9
10 /**
11 * Creates a new instance. The customer handed over the
12 * specified amount.
13 *
14 * @param paidAmt The amount of cash that was handed over
15 * by the customer.
16 */
17 public CashPayment(Amount paidAmt) {
18 this.paidAmt = paidAmt;
19 }
20
21 /**
22 * Calculates the total cost of the specified rental.
23 *
24 * @param paidRental The rental for which the customer is
25 * paying.
26 */
27 void calculateTotalCost(Rental paidRental) {
28 }
29 }

Listing C.126 Java code implementing the CashPayment class in figure 5.39

302
 Appendix C Implementations of UML Diagrams

1 package se.kth.ict.oodbook.design.casestudy.model;
2
3 /**
4 * Represents one particular rental transaction, where one
5 * particular car is rented by one particular customer.
6 */
7 public class Rental {
8 /**
9 * This rental is paid using the specified payment.
10 *
11 * @param payment The payment used to pay this rental.
12 */
13 public void pay(CashPayment payment) {
14 payment.calculateTotalCost(this);
15 }
16
17 /**
18 * Returns a receipt for the current rental.
19 */
20 public Receipt getReceipt() {
21 return new Receipt(this);
22 }
23 }

Listing C.127 Java code implementing the Rental class in figure 5.39

1 package se.kth.ict.oodbook.design.casestudy.model;
2
3 /**
4 * The receipt of a rental
5 */
6 public class Receipt {
7
8 /**
9 * Creates a new instance.
10 *
11 * @param rental The rental proved by this receipt.
12 */
13 Receipt(Rental rental) {
14 }
15
16 }

Listing C.128 Java code implementing the Receipt class in figure 5.39

303
 Appendix C Implementations of UML Diagrams

1 package se.kth.ict.oodbook.design.casestudy.integration;
2
3 import se.kth.ict.oodbook.design.casestudy.model.Receipt;
4
5 /**
6 * The interface to the printer, used for all printouts initiated
7 * by this program.
8 */
9 public class Printer {
10 public void printReceipt(Receipt receipt) {
11
12 }
13 }

Listing C.129 Java code implementing the Printer class in figure 5.39

1 package se.kth.ict.oodbook.design.casestudy.model;
2
3 /**
4 * Represents an amount of money
5 */
6 public final class Amount {
7 private final int amount;
8
9 public Amount(int amount) {
10 this.amount = amount;
11 }
12 }

Listing C.130 Java code implementing the Amount class in figure 5.39

C.29 Figure 5.40

1 package se.kth.ict.oodbook.design.casestudy.startup;
2
3 import se.kth.ict.oodbook.design.casestudy.controller.Controller;
4 import se.kth.ict.oodbook.design.casestudy.integration.Printer;
5 import se.kth.ict.oodbook.design.casestudy.integration.RegistryCreator;
6 import se.kth.ict.oodbook.design.casestudy.view.View;
7
8 /**
9 * Contains the <code>main</code> method. Performs all startup of the

304
 Appendix C Implementations of UML Diagrams

10 * application.
11 */
12 public class Main {
13 /**
14 * Starts the application.
15 *
16 * @param args The application does not take any command line
17 * parameters.
18 */
19 public static void main(String[] args) {
20 RegistryCreator creator = new RegistryCreator();
21 Printer printer = new Printer();
22 Controller contr = new Controller(creator, printer);
23 new View(contr).sampleExecution();
24 }
25 }

Listing C.131 Java code implementing the Main class in figure 5.40

1 package se.kth.ict.oodbook.design.casestudy.integration;
2
3 /**
4 * Contains all calls to the data store with cars that may be
5 * rented.
6 */
7 public class CarRegistry {
8 }

Listing C.132 Java code implementing the CarRegistry class in figure 5.40

1 package se.kth.ict.oodbook.design.casestudy.integration;
2
3 import se.kth.ict.oodbook.design.casestudy.model.Rental;
4
5 /**
6 * Contains all calls to the data store with performed rentals.
7 */
8 public class RentalRegistry {
9 }

Listing C.133 Java code implementing the RentalRegistry class in figure 5.40

1 package se.kth.ict.oodbook.design.casestudy.integration;
2

305
 Appendix C Implementations of UML Diagrams

3 /**
4 * This class is responsible for instantiating all registries.
5 */
6 public class RegistryCreator {
7 private CarRegistry carRegistry = new CarRegistry();
8 private RentalRegistry rentalRegistry = new RentalRegistry();
9
10 /**
11 * Get the value of rentalRegistry
12 *
13 * @return the value of rentalRegistry
14 */
15 public RentalRegistry getRentalRegistry() {
16 return rentalRegistry;
17 }
18
19 /**
20 * Get the value of carRegistry
21 *
22 * @return the value of carRegistry
23 */
24 public CarRegistry getCarRegistry() {
25 return carRegistry;
26 }
27 }

Listing C.134 Java code implementing the RegistryCreator class in figure 5.40

1 package se.kth.ict.oodbook.design.casestudy.integration;
2
3 /**
4 * The interface to the printer, used for all printouts
5 * initiated by this program.
6 */
7 public class Printer {
8 }

Listing C.135 Java code implementing the Printer class in figure 5.40

1 package se.kth.ict.oodbook.design.casestudy.model;
2
3 /**
4 * Represents a cash register. There shall be one
5 * instance of this class for each register.
6 */

306
 Appendix C Implementations of UML Diagrams

7 public class CashRegister {

8 }

Listing C.136 Java code implementing the CashRegister class in figure

5.40

1 package se.kth.ict.oodbook.design.casestudy.controller;
2
3 import se.kth.ict.oodbook.design.casestudy.integration.CarRegistry;
4 import se.kth.ict.oodbook.design.casestudy.integration.Printer;
5 import se.kth.ict.oodbook.design.casestudy.integration.RegistryCreator;
6 import se.kth.ict.oodbook.design.casestudy.integration.RentalRegistry;
7 import se.kth.ict.oodbook.design.casestudy.model.CashRegister;
8
9 /**
10 * This is the application’s only controller class. All calls to the
11 * model pass through here.
12 */
13 public class Controller {
14 private CarRegistry carRegistry;
15 private RentalRegistry rentalRegistry;
16 private CashRegister cashRegister;
17 private Printer printer;
18
19 /**
20 * Creates a new instance.
21 *
22 * @param regCreator Used to get all classes that handle database
23 * calls.
24 * @param printer Interface to printer.
25 */
26 public Controller(RegistryCreator regCreator, Printer printer) {
27 this.carRegistry = regCreator.getCarRegistry();
28 this.rentalRegistry = regCreator.getRentalRegistry();
29 this.printer = printer;
30 this.cashRegister = new CashRegister();
31 }
32 }

Listing C.137 Java code implementing the Controller class in figure 5.40

1 package se.kth.ict.oodbook.design.casestudy.view;
2
3 import se.kth.ict.oodbook.design.casestudy.controller.Controller;
4
5 /**

307
 Appendix C Implementations of UML Diagrams

6 * This program has no view, instead, this class is a placeholder

7 * for the entire view.
8 */
9 public class View {
10 private Controller contr;
11
12 /**
13 * Creates a new instance.
14 *
15 * @param contr The controller that is used for all operations.
16 */
17 public View(Controller contr) {
18 this.contr = contr;
19 }
20 }

Listing C.138 Java code implementing the View class in figure 5.40

C.30 Figure 5.41

1 package se.kth.ict.oodbook.design.casestudy.view;
2
3 import se.kth.ict.oodbook.design.casestudy.controller.Controller;
4
5 /**
6 * This program has no view, instead, this class is a placeholder
7 * for the entire view.
8 */
9 public class View {
10 private Controller contr;
11
12 /**
13 * Creates a new instance.
14 *
15 * @param contr The controller that is used for all
16 * operations.
17 */
18 public View(Controller contr) {
19 }
20 }

Listing C.139 Java code implementing the View class in figure 5.41

308
 Appendix C Implementations of UML Diagrams

1 package se.kth.ict.oodbook.design.casestudy.controller;
2
3 import se.kth.ict.oodbook.design.casestudy.model.Amount;
4 import se.kth.ict.oodbook.design.casestudy.integration.CarDTO;
5 import se.kth.ict.oodbook.design.casestudy.integration.CarRegistry;
6 import se.kth.ict.oodbook.design.casestudy.integration.Printer;
7 import se.kth.ict.oodbook.design.casestudy.integration.RegistryCreator;
8 import se.kth.ict.oodbook.design.casestudy.integration.RentalRegistry;
9 import se.kth.ict.oodbook.design.casestudy.model.CashPayment;
10 import se.kth.ict.oodbook.design.casestudy.model.CashRegister;
11 import se.kth.ict.oodbook.design.casestudy.model.CustomerDTO;
12 import se.kth.ict.oodbook.design.casestudy.model.Rental;
13 import se.kth.ict.oodbook.design.casestudy.model.Receipt;
14
15 /**
16 * This is the application’s only controller class. All calls to
17 * the model pass through here.
18 */
19 public class Controller {
20 private CarRegistry carRegistry;
21 private RentalRegistry rentalRegistry;
22 private Rental rental;
23
24 /**
25 * Creates a new instance.
26 *
27 * @param regCreator Used to get all classes that handle
28 * database calls.
29 */
30 public Controller(RegistryCreator regCreator) {
31 }
32
33 /**
34 * Search for a car matching the specified search criteria.
35 *
36 * @param searchedCar This object contains the search criteria.
37 * Fields in the object that are set to
38 * <code>null</code> or
39 * <code>false</code> are ignored.
40 * @return The best match of the search criteria.
41 */
42 public CarDTO searchMatchingCar(CarDTO searchedCar) {
43 }
44
45 /**
46 * Registers a new customer. Only registered customers can

309
 Appendix C Implementations of UML Diagrams

47 * rent cars.
48 *
49 * @param customer The customer that will be registered.
50 */
51 public void registerCustomer(CustomerDTO customer) {
52 }
53
54 /**
55 * Books the specified car. After calling this method, the car
56 * can not be booked by any other customer. This method also
57 * permanently saves information about the current rental.
58 *
59 * @param car The car that will be booked.
60 */
61 public void bookCar(CarDTO car) {
62 }
63
64 /**
65 * Handles rental payment. Updates the balance of the cash register
66 * where the payment was performed. Calculates change. Prints the
67 * receipt.
68 *
69 * @param paidAmt The paid amount.
70 */
71 public void pay(Amount paidAmt) {
72 }
73 }

Listing C.140 Java code implementing the Controller class in figure 5.41

1 package se.kth.ict.oodbook.design.casestudy.startup;
2
3 import se.kth.ict.oodbook.design.casestudy.controller.Controller;
4 import se.kth.ict.oodbook.design.casestudy.integration.Printer;
5 import se.kth.ict.oodbook.design.casestudy.integration.RegistryCreator;
6 import se.kth.ict.oodbook.design.casestudy.view.View;
7
8 /**
9 * Contains the <code>main</code> method. Performs all startup of
10 * the application.
11 */
12 public class Main {
13 /**
14 * Starts the application.
15 *
16 * @param args The application does not take any command line

310
 Appendix C Implementations of UML Diagrams

17 * parameters.
18 */
19 public static void main(String[] args) {
20 }
21 }

Listing C.141 Java code implementing the Main class in figure 5.41

1 package se.kth.ict.oodbook.design.casestudy.model;
2
3 import se.kth.ict.oodbook.design.casestudy.integration.CarDTO;
4 import se.kth.ict.oodbook.design.casestudy.integration.CarRegistry;
5
6 /**
7 * Represents one particular rental transaction, where one
8 * particular car is rented by one particular customer.
9 */
10 public class Rental {
11 private CarRegistry carRegistry;
12
13 /**
14 * Creates a new instance, representing a rental made by the
15 * specified customer.
16 *
17 * @param customer The renting customer.
18 * @param carRegistry The data store with information about
19 * available cars.
20 */
21 public Rental(CustomerDTO customer, CarRegistry carRegistry) {
22 }
23
24 /**
25 * Specifies the car that was rented.
26 *
27 * @param rentedCar The car that was rented.
28 */
29 public void setRentedCar(CarDTO rentedCar) {
30 }
31
32 /**
33 * This rental is paid using the specified payment.
34 *
35 * @param payment The payment used to pay this rental.
36 */
37 public void pay(CashPayment payment) {
38 }

311
 Appendix C Implementations of UML Diagrams

39
40 /**
41 * Returns a receipt for the current rental.
42 */
43 public Receipt getReceipt() {
44 }
45 }

Listing C.142 Java code implementing the Rental class in figure 5.41

1 package se.kth.ict.oodbook.design.casestudy.model;
2
3 /**
4 * The receipt of a rental
5 */
6 public class Receipt {
7 /**
8 * Creates a new instance.
9 *
10 * @param rental The rental proved by this receipt.
11 */
12 Receipt(Rental rental) {
13 }
14 }

Listing C.143 Java code implementing the Receipt class in figure 5.41

1 package se.kth.ict.oodbook.design.casestudy.model;
2
3 /**
4 * Represents one specific payment for one specific rental.
5 * The rental is payed with cash.
6 */
7 public class CashPayment {
8 /**
9 * Creates a new instance. The customer handed over
10 * the specified amount.
11 *
12 * @param paidAmt The amount of cash that was handed
13 * over by the customer.
14 */
15 public CashPayment(Amount paidAmt) {
16 }
17
18 /**

312
 Appendix C Implementations of UML Diagrams

19 * Calculates the total cost of the specified rental.

20 *
21 * @param paidRental The rental for which the customer
22 * is paying.
23 */
24 void calculateTotalCost(Rental paidRental) {
25 }
26 }

Listing C.144 Java code implementing the CashPayment class in figure 5.41

1 package se.kth.ict.oodbook.design.casestudy.model;
2
3 /**
4 * Represents a cash register. There shall be one instance
5 * of this class for each register.
6 */
7 public class CashRegister {
8 public void addPayment(CashPayment payment) {
9 }
10 }

Listing C.145 Java code implementing the CashRegister class in figure 5.41

1 package se.kth.ict.oodbook.design.casestudy.integration;
2
3 import se.kth.ict.oodbook.design.casestudy.model.Receipt;
4
5 /**
6 * The interface to the printer, used for all printouts
7 * initiated by this program.
8 */
9 public class Printer {
10 public void printReceipt(Receipt receipt) {
11 }
12 }

Listing C.146 Java code implementing the Printer class in figure 5.41

1 package se.kth.ict.oodbook.design.casestudy.integration;
2
3 /**
4 * This class is responsible for instantiating all registries.
5 */

313
 Appendix C Implementations of UML Diagrams

6 public class RegistryCreator {

7 /**
8 * Get the value of rentalRegistry
9 *
10 * @return the value of rentalRegistry
11 */
12 public RentalRegistry getRentalRegistry() {
13 }
14
15 /**
16 * Get the value of carRegistry
17 *
18 * @return the value of carRegistry
19 */
20 public CarRegistry getCarRegistry() {
21 }
22 }

Listing C.147 Java code implementing the RegistryCreator class in figure 5.41

1 package se.kth.ict.oodbook.design.casestudy.integration;
2
3 /**
4 * Contains all calls to the data store with cars that may be
5 * rented.
6 */
7 public class CarRegistry {
8 CarRegistry() {
9 }
10
11 /**
12 * Search for a car matching the specified search criteria.
13 *
14 * @param searchedCar This object contains the search criteria.
15 * Fields in the object that are set to
16 * <code>null</code> or <code>false</code>
17 * are ignored.
18 * @return The best match of the search criteria.
19 */
20 public CarDTO findCar(CarDTO searchedCar) {
21 }
22
23 /**
24 * Books the specified car. After calling this method, the car
25 * can not be booked by any other customer.
26 *

314
 Appendix C Implementations of UML Diagrams

27 * @param car The car that will be booked.

28 */
29 public void bookCar(CarDTO car) {
30 }
31 }

Listing C.148 Java code implementing the CarRegistry class in figure 5.41

1 package se.kth.ict.oodbook.design.casestudy.integration;
2
3 import se.kth.ict.oodbook.design.casestudy.model.Rental;
4
5 /**
6 * Contains all calls to the data store with performed rentals.
7 */
8 public class RentalRegistry {
9 RentalRegistry() {
10 }
11
12 /**
13 * Saves the specified rental permanently.
14 *
15 * @param rental The rental that will be saved.
16 */
17 public void saveRental(Rental rental) {
18 }
19 }

Listing C.149 Java code implementing the RentalRegistry class in figure 5.41

C.31 Figure 5.42

1 package se.kth.ict.oodbook.design.casestudy.view;
2
3 import se.kth.ict.oodbook.design.casestudy.controller.Controller;
4 import se.kth.ict.oodbook.design.casestudy.model.Amount;
5
6 /**
7 * This program has no view, instead, this class is a placeholder
8 * for the entire view.
9 */
10 public class View {
11 private Controller contr;

315
 Appendix C Implementations of UML Diagrams

12
13 // Somewhere in the code. The used amount (100) is not
14 // specified in the diagram.
15 Amount paidAmount = new Amount(100);
16 contr.pay(paidAmount);
17 }

Listing C.150 Java code implementing the View class in figure 5.42

1 package se.kth.ict.oodbook.design.casestudy.controller;
2
3 import se.kth.ict.oodbook.design.casestudy.model.Amount;
4 import se.kth.ict.oodbook.design.casestudy.integration.Printer;
5 import se.kth.ict.oodbook.design.casestudy.model.CashPayment;
6 import se.kth.ict.oodbook.design.casestudy.model.CashRegister;
7 import se.kth.ict.oodbook.design.casestudy.model.Rental;
8
9 /**
10 * This is the application’s only controller class. All
11 * calls to the model pass through here.
12 */
13 public class Controller {
14 private Rental rental;
15 private CashRegister cashRegister;
16 private Printer printer;
17
18 /**
19 * Handles rental payment. Updates the balance of the cash
20 * register where the payment was performed. Calculates
21 * change. Prints the receipt.
22 *
23 * @param paidAmt The paid amount.
24 */
25 public void pay(Amount paidAmt) {
26 CashPayment payment = new CashPayment(paidAmt);
27 rental.pay(payment);
28 cashRegister.addPayment(payment);
29 rental.printReceipt(printer);
30 }
31 }

Listing C.151 Java code implementing the Controller class in figure 5.42

1 package se.kth.ict.oodbook.design.casestudy.model;
2

316
 Appendix C Implementations of UML Diagrams

3 /**
4 * Represents a cash register. There shall be one instance of
5 * this class for each register.
6 */
7 public class CashRegister {
8 public void addPayment(CashPayment payment) {
9 }
10 }

Listing C.152 Java code implementing the CashRegister class in figure 5.42

1 package se.kth.ict.oodbook.design.casestudy.model;
2
3 /**
4 * Represents one specific payment for one specific rental. The
5 * rental is payed with cash.
6 */
7 public class CashPayment {
8 private Amount paidAmt;
9
10 /**
11 * Creates a new instance. The customer handed over the
12 * specified amount.
13 *
14 * @param paidAmt The amount of cash that was handed over
15 * by the customer.
16 */
17 public CashPayment(Amount paidAmt) {
18 this.paidAmt = paidAmt;
19 }
20
21 /**
22 * Calculates the total cost of the specified rental.
23 *
24 * @param paidRental The rental for which the customer is
25 * paying.
26 */
27 void calculateTotalCost(Rental paidRental) {
28 }
29 }

Listing C.153 Java code implementing the CashPayment class in figure 5.42

1 package se.kth.ict.oodbook.design.casestudy.model;
2

317
 Appendix C Implementations of UML Diagrams

3 import se.kth.ict.oodbook.design.casestudy.integration.Printer;
4
5 /**
6 * Represents one particular rental transaction, where one
7 * particular car is rented by one particular customer.
8 */
9 public class Rental {
10 /**
11 * This rental is paid using the specified payment.
12 *
13 * @param payment The payment used to pay this rental.
14 */
15 public void pay(CashPayment payment) {
16 payment.calculateTotalCost(this);
17 }
18
19 /**
20 * Prints a receipt for the current rental on the
21 * specified printer.
22 */
23 public void printReceipt(Printer printer) {
24 Receipt receipt = new Receipt(this);
25 printer.printReceipt(receipt);
26 }
27 }

Listing C.154 Java code implementing the Rental class in figure 5.42

1 package se.kth.ict.oodbook.design.casestudy.model;
2
3 /**
4 * The receipt of a rental
5 */
6 public class Receipt {
7
8 /**
9 * Creates a new instance.
10 *
11 * @param rental The rental proved by this receipt.
12 */
13 Receipt(Rental rental) {
14 }
15
16 }

Listing C.155 Java code implementing the Receipt class in figure 5.42

318
 Appendix C Implementations of UML Diagrams

1 package se.kth.ict.oodbook.design.casestudy.integration;
2
3 import se.kth.ict.oodbook.design.casestudy.model.Receipt;
4
5 /**
6 * The interface to the printer, used for all printouts initiated
7 * by this program.
8 */
9 public class Printer {
10 public void printReceipt(Receipt receipt) {
11
12 }
13 }

Listing C.156 Java code implementing the Printer class in figure 5.42

1 package se.kth.ict.oodbook.design.casestudy.model;
2
3 /**
4 * Represents an amount of money
5 */
6 public final class Amount {
7 private final int amount;
8
9 public Amount(int amount) {
10 this.amount = amount;
11 }
12 }

Listing C.157 Java code implementing the Amount class in figure 5.42

C.32 Figure 5.43

1 package se.kth.ict.oodbook.design.casestudy.view;
2
3 import se.kth.ict.oodbook.design.casestudy.controller.Controller;
4
5 /**
6 * This program has no view, instead, this class is a placeholder
7 * for the entire view.
8 */
9 public class View {

319
 Appendix C Implementations of UML Diagrams

10 private Controller contr;

11
12 /**
13 * Creates a new instance.
14 *
15 * @param contr The controller that is used for all
16 * operations.
17 */
18 public View(Controller contr) {
19 }
20 }

Listing C.158 Java code implementing the View class in figure 5.43

1 package se.kth.ict.oodbook.design.casestudy.controller;
2
3 import se.kth.ict.oodbook.design.casestudy.model.Amount;
4 import se.kth.ict.oodbook.design.casestudy.integration.CarDTO;
5 import se.kth.ict.oodbook.design.casestudy.integration.CarRegistry;
6 import se.kth.ict.oodbook.design.casestudy.integration.Printer;
7 import se.kth.ict.oodbook.design.casestudy.integration.RegistryCreator;
8 import se.kth.ict.oodbook.design.casestudy.integration.RentalRegistry;
9 import se.kth.ict.oodbook.design.casestudy.model.CashPayment;
10 import se.kth.ict.oodbook.design.casestudy.model.CashRegister;
11 import se.kth.ict.oodbook.design.casestudy.model.CustomerDTO;
12 import se.kth.ict.oodbook.design.casestudy.model.Rental;
13 import se.kth.ict.oodbook.design.casestudy.model.Receipt;
14
15 /**
16 * This is the application’s only controller class. All calls to
17 * the model pass through here.
18 */
19 public class Controller {
20 private CarRegistry carRegistry;
21 private RentalRegistry rentalRegistry;
22 private Rental rental;
23
24 /**
25 * Creates a new instance.
26 *
27 * @param regCreator Used to get all classes that handle
28 * database calls.
29 */
30 public Controller(RegistryCreator regCreator) {
31 }
32

320
 Appendix C Implementations of UML Diagrams

33 /**
34 * Search for a car matching the specified search criteria.
35 *
36 * @param searchedCar This object contains the search criteria.
37 * Fields in the object that are set to
38 * <code>null</code> or
39 * <code>false</code> are ignored.
40 * @return The best match of the search criteria.
41 */
42 public CarDTO searchMatchingCar(CarDTO searchedCar) {
43 }
44
45 /**
46 * Registers a new customer. Only registered customers can
47 * rent cars.
48 *
49 * @param customer The customer that will be registered.
50 */
51 public void registerCustomer(CustomerDTO customer) {
52 }
53
54 /**
55 * Books the specified car. After calling this method, the car
56 * can not be booked by any other customer. This method also
57 * permanently saves information about the current rental.
58 *
59 * @param car The car that will be booked.
60 */
61 public void bookCar(CarDTO car) {
62 }
63
64 /**
65 * Handles rental payment. Updates the balance of the cash register
66 * where the payment was performed. Calculates change. Prints the
67 * receipt.
68 *
69 * @param paidAmt The paid amount.
70 */
71 public void pay(Amount paidAmt) {
72 }
73 }

Listing C.159 Java code implementing the Controller class in figure 5.43

1 package se.kth.ict.oodbook.design.casestudy.startup;
2

321
 Appendix C Implementations of UML Diagrams

3 import se.kth.ict.oodbook.design.casestudy.controller.Controller;
4 import se.kth.ict.oodbook.design.casestudy.integration.Printer;
5 import se.kth.ict.oodbook.design.casestudy.integration.RegistryCreator;
6 import se.kth.ict.oodbook.design.casestudy.view.View;
7
8 /**
9 * Contains the <code>main</code> method. Performs all startup of
10 * the application.
11 */
12 public class Main {
13 /**
14 * Starts the application.
15 *
16 * @param args The application does not take any command line
17 * parameters.
18 */
19 public static void main(String[] args) {
20 }
21 }

Listing C.160 Java code implementing the Main class in figure 5.43

1 package se.kth.ict.oodbook.design.casestudy.model;
2
3 import se.kth.ict.oodbook.design.casestudy.integration.CarDTO;
4 import se.kth.ict.oodbook.design.casestudy.integration.CarRegistry;
5 import se.kth.ict.oodbook.design.casestudy.integration.Printer;
6
7 /**
8 * Represents one particular rental transaction, where one
9 * particular car is rented by one particular customer.
10 */
11 public class Rental {
12 private CarRegistry carRegistry;
13
14 /**
15 * Creates a new instance, representing a rental made by the
16 * specified customer.
17 *
18 * @param customer The renting customer.
19 * @param carRegistry The data store with information about
20 * available cars.
21 */
22 public Rental(CustomerDTO customer, CarRegistry carRegistry) {
23 }
24

322
 Appendix C Implementations of UML Diagrams

25 /**
26 * Specifies the car that was rented.
27 *
28 * @param rentedCar The car that was rented.
29 */
30 public void setRentedCar(CarDTO rentedCar) {
31 }
32
33 /**
34 * This rental is paid using the specified payment.
35 *
36 * @param payment The payment used to pay this rental.
37 */
38 public void pay(CashPayment payment) {
39 }
40
41 /**
42 * Prints a receipt for the current rental on the specified
43 * printer.
44 */
45 public void printReceipt(Printer printer) {
46 }
47 }

Listing C.161 Java code implementing the Rental class in figure 5.43

1 package se.kth.ict.oodbook.design.casestudy.model;
2
3 /**
4 * The receipt of a rental
5 */
6 public class Receipt {
7 /**
8 * Creates a new instance.
9 *
10 * @param rental The rental proved by this receipt.
11 */
12 Receipt(Rental rental) {
13 }
14 }

Listing C.162 Java code implementing the Receipt class in figure 5.43

1 package se.kth.ict.oodbook.design.casestudy.model;
2

323
 Appendix C Implementations of UML Diagrams

3 /**
4 * Represents one specific payment for one specific rental.
5 * The rental is payed with cash.
6 */
7 public class CashPayment {
8 /**
9 * Creates a new instance. The customer handed over
10 * the specified amount.
11 *
12 * @param paidAmt The amount of cash that was handed
13 * over by the customer.
14 */
15 public CashPayment(Amount paidAmt) {
16 }
17
18 /**
19 * Calculates the total cost of the specified rental.
20 *
21 * @param paidRental The rental for which the customer
22 * is paying.
23 */
24 void calculateTotalCost(Rental paidRental) {
25 }
26 }

Listing C.163 Java code implementing the CashPayment class in figure 5.43

1 package se.kth.ict.oodbook.design.casestudy.model;
2
3 /**
4 * Represents a cash register. There shall be one instance
5 * of this class for each register.
6 */
7 public class CashRegister {
8 public void addPayment(CashPayment payment) {
9 }
10 }

Listing C.164 Java code implementing the CashRegister class in figure 5.43

1 package se.kth.ict.oodbook.design.casestudy.integration;
2
3 import se.kth.ict.oodbook.design.casestudy.model.Receipt;
4
5 /**

324
 Appendix C Implementations of UML Diagrams

6 * The interface to the printer, used for all printouts

7 * initiated by this program.
8 */
9 public class Printer {
10 public void printReceipt(Receipt receipt) {
11 }
12 }

Listing C.165 Java code implementing the Printer class in figure 5.43

1 package se.kth.ict.oodbook.design.casestudy.integration;
2
3 /**
4 * This class is responsible for instantiating all registries.
5 */
6 public class RegistryCreator {
7 /**
8 * Get the value of rentalRegistry
9 *
10 * @return the value of rentalRegistry
11 */
12 public RentalRegistry getRentalRegistry() {
13 }
14
15 /**
16 * Get the value of carRegistry
17 *
18 * @return the value of carRegistry
19 */
20 public CarRegistry getCarRegistry() {
21 }
22 }

Listing C.166 Java code implementing the RegistryCreator class in figure 5.43

1 package se.kth.ict.oodbook.design.casestudy.integration;
2
3 /**
4 * Contains all calls to the data store with cars that may be
5 * rented.
6 */
7 public class CarRegistry {
8 CarRegistry() {
9 }
10

325
 Appendix C Implementations of UML Diagrams

11 /**
12 * Search for a car matching the specified search criteria.
13 *
14 * @param searchedCar This object contains the search criteria.
15 * Fields in the object that are set to
16 * <code>null</code> or <code>false</code>
17 * are ignored.
18 * @return The best match of the search criteria.
19 */
20 public CarDTO findCar(CarDTO searchedCar) {
21 }
22
23 /**
24 * Books the specified car. After calling this method, the car
25 * can not be booked by any other customer.
26 *
27 * @param car The car that will be booked.
28 */
29 public void bookCar(CarDTO car) {
30 }
31 }

Listing C.167 Java code implementing the CarRegistry class in figure 5.43

1 package se.kth.ict.oodbook.design.casestudy.integration;
2
3 import se.kth.ict.oodbook.design.casestudy.model.Rental;
4
5 /**
6 * Contains all calls to the data store with performed rentals.
7 */
8 public class RentalRegistry {
9 RentalRegistry() {
10 }
11
12 /**
13 * Saves the specified rental permanently.
14 *
15 * @param rental The rental that will be saved.
16 */
17 public void saveRental(Rental rental) {
18 }
19 }

Listing C.168 Java code implementing the RentalRegistry class in figure 5.43

326
 Appendix C Implementations of UML Diagrams

C.33 Figure 8.2

Package names are not shown in the diagram, but have been added in the code.

1 package se.kth.ict.oodbook.exception.uml;
2
3 public class MyClass {
4 public void myMethod() throws MyException, AnotherException {
5 }
6 }

Listing C.169 Java code implementing the MyClass class in figure 8.2a

1 package se.kth.ict.oodbook.exception.uml;
2
3 public class MyException extends Exception {
4 }

Listing C.170 Java code implementing the MyException class

in figure 8.2a

1 package se.kth.ict.oodbook.exception.uml;
2
3 public class AnotherException extends Exception {
4 }

Listing C.171 Java code implementing the AnotherException class in

figure 8.2a

1 package se.kth.ict.oodbook.exception.uml;
2
3 public class MyClass {
4 // The diagram shows that one or more of the methods in
5 // this class throws MyException, but it does not show
6 // which method(s).
7 public void myMethod() {
8 }
9
10 public void anotherMethod() {
11 }
12 }

Listing C.172 Java code implementing the MyClass class in figure 8.2b

1 package se.kth.ict.oodbook.exception.uml;

327
 Appendix C Implementations of UML Diagrams

2
3 public class MyException extends Exception {
4 }

Listing C.173 Java code implementing the MyException class

in figure 8.2b

C.34 Figure 8.3

Package names are not shown in the diagram, but have been added in the code. The code is
exactly the same for figures 8.3a and 8.3b.

1 package se.kth.ict.oodbook.exception.uml;
2
3 public class MyClass {
4 public void myMethod() throws MyException {
5 }
6 }

Listing C.174 Java code implementing the MyClass class in figure 8.3

1 package se.kth.ict.oodbook.exception.uml;
2
3 public class MyException extends Exception {
4 }

Listing C.175 Java code implementing the MyException class

in figure 8.3

1 package se.kth.ict.oodbook.exception.uml;
2
3 public class SomeClass {
4 private MyClass myClass;
5
6 // Somewhere, in some method in this class:
7 try {
8 myClass.myMethod();
9 } catch (MyException exception) {
10
11 }
12 }

Listing C.176 Java code implementing the SomeClass class in figure 8.3.
Strictly speaking, the UML diagram does not tell whether there is a try-catch
block or not.

328
 Appendix C Implementations of UML Diagrams

C.35 Figure 9.1

Package names are not shown in the diagram, but have been added in the code.

1 package se.leiflindback.oodbook.polymorphism.uml;
2
3 interface InterfaceA {
4 public void methodA();
5 }

Listing C.177 Java code implementing the InterfaceA interface in fig-

ure 9.1a

1 package se.leiflindback.oodbook.polymorphism.uml;
2
3 public class ClassA implements InterfaceA {
4 @Override
5 public void methodA() {
6 }
7 }

Listing C.178 Java code implementing the ClassA class in figure 9.1a

1 package se.leiflindback.oodbook.polymorphism.uml;
2
3 public class ClassB extends ClassC{
4 }

Listing C.179 Java code implementing the ClassB class in figure 9.1b

1 package se.leiflindback.oodbook.polymorphism.uml;
2
3 public class ClassC {
4 }

Listing C.180 Java code implementing the ClassC class in figure 9.1b

C.36 Figure 9.2

Package names are not shown in the diagram, but have been added in the code.

1 package se.leiflindback.oodbook.polymorphism.uml.seqClass;
2

329
 Appendix C Implementations of UML Diagrams

3 public class AnyClass {

4 private ClassA callee;
5
6 //Somewhere in some method.
7 callee.methodA();
8 }

Listing C.181 Java code implementing the AnyClass object in figure 9.2a

1 package se.leiflindback.oodbook.polymorphism.uml.seqClass;
2
3 public class ClassA {
4 public void methodA() {
5 }
6 }

Listing C.182 Java code implementing the ClassA object in figure 9.2a

1 package se.leiflindback.oodbook.polymorphism.uml.seqInterf;
2
3 public class AnyClass {
4 private InterfaceA callee;
5
6 //Somewhere in some method.
7 callee.methodA();
8 }

Listing C.183 Java code implementing the AnyClass object in figure 9.2b

1 package se.leiflindback.oodbook.polymorphism.uml.seqInterf;
2
3 interface InterfaceA {
4 public void methodA();
5 }

Listing C.184 Java code implementing the InterfaceA object in figure 9.2b

1 package se.leiflindback.oodbook.polymorphism.uml.seqInterfClass;
2
3 public class AnyClass {
4 private InterfaceA callee;
5
6 //Somewhere in some method.
7 callee.methodA();

330
 Appendix C Implementations of UML Diagrams

8 }

Listing C.185 Java code implementing the AnyClass object in figure 9.2c

1 package se.leiflindback.oodbook.polymorphism.uml.seqInterfClass;
2
3 public class SomeUnknownClass implements InterfaceA {
4 private ClassA callee;
5
6 @Override
7 public void methodA() {
8 callee.methodA();
9 }
10 }

Listing C.186 Java code implementing the InterfaceA object in figure 9.2c

1 package se.leiflindback.oodbook.polymorphism.uml.seqInterfClass;
2
3 public class ClassA {
4 public void methodA() {
5 }
6 }

Listing C.187 Java code implementing the ClassA object in figure 9.2c

C.37 Figure 9.3

The implementation of this diagram identical to that of diagram 9.2b above (listings C.183
and C.184), since these two figures are different ways to show exactly the same thing.

C.38 Figure 9.6

1 package package1;
2
3 public class Class1 {
4 protected void protectedMethod() {
5
6 }
7 }

331
 Appendix C Implementations of UML Diagrams

Listing C.188 Java code implementing Class1 in

figure 9.6a.

1 package package1;
2
3 public class Class2 extends Class1 {
4
5 }

Listing C.189 Java code implementing Class2 in

figure 9.6a.

1 package package1;
2
3 public class Class1 {
4 protected void protectedMethod() {
5
6 }
7 }

Listing C.190 Java code implementing Class1 in

figure 9.6b.

1 package package1;
2
3 public class Class2 {
4 private Class1 class1;
5 }

Listing C.191 Java code imple-

menting Class2 in figure 9.6b.

1 package package1;
2
3 public class Class1 {
4 protected void protectedMethod() {
5
6 }
7 }

Listing C.192 Java code implementing Class1 in

figure 9.6c.

1 package package2;

332
 Appendix C Implementations of UML Diagrams

2
3 import package1.Class1;
4
5 public class Class2 extends Class1 {
6
7 }

Listing C.193 Java code implementing Class2 in

figure 9.6c.

1 package package1;
2
3 public class Class1 {
4 protected void protectedMethod() {
5
6 }
7 }

Listing C.194 Java code implementing Class1 in

figure 9.6d.

1 package package2;
2
3 import package1.Class1;
4
5 public class Class2 {
6 private Class1 class1;
7 }

Listing C.195 Java code imple-

menting Class2 in figure 9.6d.

C.39 Figure 9.7

1 public class Animal {

2 private String name;
3 }
4
5 public class Bird extends Animal {
6 public void fly() {
7 }
8 }
9
10 public class Fish extends Animal {

333
 Appendix C Implementations of UML Diagrams

11 public void swim() {

12 }
13 }
14
15 public class Mammal extends Animal {
16 public void walk() {
17 }
18 }
19
20 public class Crow extends Bird {
21 }
22
23 public class Parrot extends Bird {
24 }
25
26 public class Salmon extends Fish {
27 }
28
29 public class Perch extends Fish {
30 }
31
32 public class Dog extends Mammal {
33 }
34
35 public class Cat extends Mammal {
36 }

Listing C.196 Java code implementing all classes in figure 9.7.

C.40 Figure 9.8

1 public class Animal {

2 private String name;
3 }
4
5 public class Flyer extends Animal {
6 public void fly() {
7 }
8 }
9
10 public class Swimmer extends Animal {
11 public void swim() {
12 }
13 }

334
 Appendix C Implementations of UML Diagrams

14
15 public class Walker extends Animal {
16 public void walk() {
17 }
18 }
19
20 public class Parrot extends Flyer {
21 }
22
23 public class Crow extends Flyer {
24 }
25
26 public class Salmon extends Swimmer {
27 }
28
29 public class Perch extends Swimmer {
30 }
31
32 public class Penguin extends Swimmer {
33 }
34
35 public class Cat extends Walker {
36 }
37
38 public class Dog extends Walker {
39 }

Listing C.197 Java code implementing all classes in figure 9.8.

C.41 Figure 9.9

1 public class Animal {

2 private String name;
3 }
4
5 public class Flyer {
6 public void fly() {
7 }
8 }
9
10 public class Swimmer {
11 public void swim() {
12 }
13 }

335
 Appendix C Implementations of UML Diagrams

14
15 public class Walker {
16 public void walk() {
17 }
18 }
19
20 public class Parrot {
21 private Animal animal;
22 private Flyer flyer;
23
24 public void fly() {
25 flyer.fly();
26 }
27 }
28
29 public class Crow {
30 private Animal animal;
31 private Flyer flyer;
32
33 public void fly() {
34 flyer.fly();
35 }
36 }
37
38 public class Salmon {
39 private Animal animal;
40 private Swimmer swimmer;
41
42 public void swim() {
43 swimmer.swim();
44 }
45 }
46
47 public class Perch {
48 private Animal animal;
49 private Swimmer swimmer;
50
51 public void swim() {
52 swimmer.swim();
53 }
54 }
55
56 public class Ostrich {
57 private Animal animal;
58 private Walker walker;
59

336
 Appendix C Implementations of UML Diagrams

60 public void walk() {

61 walker.walk();
62 }
63 }
64
65 public class Cat {
66 private Animal animal;
67 private Walker walker;
68 private Swimmer swimmer;
69
70 public void walk() {
71 walker.walk();
72 }
73
74 public void swim() {
75 swimmer.swim();
76 }
77 }

Listing C.198 Java code implementing all classes in figure 9.9.

C.42 Figure 9.12

Package names are not shown in the diagram, but have been added in the code.

1 package se.leiflindback.oodbook.despat.observer;
2
3 import java.util.ArrayList;
4 import java.util.List;
5
6 /**
7 * The observed class in a general implementation of the
8 * observer pattern.
9 */
10 public class ObservedClass {
11 private List<Observer> observers = new ArrayList<>();
12
13 /**
14 * Registers observers. Any <code>Observer</code> that is
15 * passed to this method will be notified when this object
16 * changes state.
17 *
18 * @param observer The observer that shall be registered.
19 */
20 public void addObserver(Observer observer) {

337
 Appendix C Implementations of UML Diagrams

21 observers.add(observer);
22 }
23
24 // Called by any method in this class that has changed the
25 // class’ state.
26 private void notifyObservers() {
27 for (Observer observer : observers) {
28 observer.stateHasChanged();
29 }
30 }
31 }

Listing C.199 Java code implementing ObservedClass in figure 9.12. The method
notifyObservers is not shown in the diagram.

1 package se.leiflindback.oodbook.despat.observer;
2
3 /**
4 * The observer interface in a general implementation of the
5 * observer pattern.
6 */
7 public interface Observer {
8 /**
9 * Called when the observed class changes state.
10 */
11 void stateHasChanged();
12 }

Listing C.200 Java code implementing Observer in figure 9.12.

1 package se.leiflindback.oodbook.despat.observer;
2
3 /**
4 * The observing class in a general implementation of the
5 * observer pattern.
6 */
7 public class AnyClassThatImplementsObserver implements
8 Observer {
9 @Override
10 public void stateHasChanged() {
11 }
12 }

Listing C.201 Java code implementing AnyClassThatImplementsObserver in fig-

ure 9.12.

338
 Appendix C Implementations of UML Diagrams

C.43 Figure 9.14

Package names are not shown in the diagram, but have been added in the code.

1 package se.leiflindback.oodbook.polymInherit.strategy;
2
3 /**
4 * A class that uses a concrete strategy.
5 */
6 public class Client {
7 private StrategyDefinition strategy;
8 }

Listing C.202 Java code implementing Client in figure 9.14.

1 package se.leiflindback.oodbook.polymInherit.strategy;
2
3 /**
4 * A general example of definition of a strategy.
5 */
6 public interface StrategyDefinition {
7 /**
8 * The work performed by this strategy.
9 */
10 public void algorithm();
11 }

Listing C.203 Java code implementing StrategyDefinition in figure 9.14.

1 package se.leiflindback.oodbook.polymInherit.strategy;
2
3 /**
4 * A concrete implementation of a strategy.
5 */
6 public class ConcreteStrategyA implements StrategyDefinition {
7 @Override
8 public void algorithm() {
9 }
10 }

Listing C.204 Java code implementing ConcreteStrategyA in figure 9.14.

1 package se.leiflindback.oodbook.polymInherit.strategy;
2

339
 Appendix C Implementations of UML Diagrams

3 /**
4 * A concrete implementation of a strategy.
5 */
6 public class ConcreteStrategyB implements StrategyDefinition {
7 @Override
8 public void algorithm() {
9 }
10 }

Listing C.205 Java code implementing ConcreteStrategyB in figure 9.14.

C.44 Figure 9.17

Package names are not shown in the diagram, but have been added in the code.

1 package se.leiflindback.oodbook.polymInherit.factory;
2
3 /**
4 * A class that uses a concrete strategy.
5 */
6 public class Client {
7 private Product product = new Factory().createProduct();
8 }

Listing C.206 Java code implementing Client in figure 9.17.

1 package se.leiflindback.oodbook.polymInherit.factory;
2
3 /**
4 * A generic example of a definition of products created by
5 * a factory.
6 */
7 public interface Product {
8 }

Listing C.207 Java code implementing Product in figure 9.17.

1 package se.leiflindback.oodbook.polymInherit.factory;
2
3 /**
4 * An example of a concrete product
5 */
6 public class ConcreteProductA implements Product {

340
 Appendix C Implementations of UML Diagrams

7 }

Listing C.208 Java code implementing ProductA in figure 9.17.

1 package se.leiflindback.oodbook.polymInherit.factory;
2
3 /**
4 * An example of a concrete product
5 */
6 public class ConcreteProductB implements Product {
7 }

Listing C.209 Java code implementing ProductB in figure 9.17.

1 package se.leiflindback.oodbook.polymInherit.factory;
2
3 /**
4 * A generic example of a factory.
5 */
6 public class Factory {
7
8 /**
9 * Creates a new product.
10 * @return The newly created product.
11 */
12 public Product createProduct() {
13 return new ConcreteProductA();
14 }
15 }

Listing C.210 Java code implementing Factory in figure 9.17.

C.45 Figures 9.21 and 9.22

Diagrams 9.21 and 9.22 illustrate exactly the same code. Package names are not shown in the
diagrams, but have been added in the code.

1 package se.leiflindback.oodbook.polyminherit.composite;
2
3 /**
4 * The client of the algorithm.
5 */
6 public class Client {

341
 Appendix C Implementations of UML Diagrams

7 private Task task;

8
9 public void anyMethodInThisClass() {
10 // Task shall be a reference to Composite, but the
11 // diagrams do not show how that reference is created.
12 task.performTask();
13 }
14 }

Listing C.211 Java code implementing Client in figurs 9.21 and 9.22.

1 package se.leiflindback.oodbook.polyminherit.composite;
2
3 /**
4 * A definition of an algorithm.
5 */
6 public interface Task {
7 /**
8 * Performs the algorithm.
9 */
10 void performTask();
11 }

Listing C.212 Java code implementing Task in figurs 9.21 and 9.22.

1 package se.leiflindback.oodbook.polyminherit.composite;
2
3 /**
4 * An implementation of the algorithm.
5 */
6 public class ConcreteTaskA implements Task {
7 @Override
8 public void performTask() {
9 }
10 }

Listing C.213 Java code implementing ConcreteTaskA in figurs 9.21 and 9.22.

1 package se.leiflindback.oodbook.polyminherit.composite;
2
3 /**
4 * An implementation of the algorithm.
5 */
6 public class ConcreteTaskB implements Task {

342
 Appendix C Implementations of UML Diagrams

7 @Override
8 public void performTask() {
9 }
10 }

Listing C.214 Java code implementing ConcreteTaskB in figurs 9.21 and 9.22.

1 package se.leiflindback.oodbook.polyminherit.composite;
2
3 import java.util.List;
4
5 /**
6 * A composite algorithm, containing concrete implementations
7 * of the algorithm.
8 */
9 public class Composite implements Task {
10 // The diagrams do not show how concrete tasks are added
11 // to the list.
12 private List<Task> tasks;
13
14 @Override
15 public void performTask() {
16 for (Task task : tasks) {
17 task.performTask();
18 }
19 }
20 }

Listing C.215 Java code implementing Composite in figurs 9.21 and 9.22.

C.46 Figure 9.24

Diagrams 9.24a and 9.24b illustrate exactly the same code. Package names are not shown in
the diagrams, but have been added in the code.

1 package se.leiflindback.oodbook.polyminherit.tempmet;
2
3 /**
4 * A class that uses a concrete implementation of a template
5 * method.
6 */
7 public class Client {
8 private TaskTemplate task;
9 public void anyMethod() {

343
 Appendix C Implementations of UML Diagrams

10 task.performTask();
11 }
12 }

Listing C.216 Java code implementing Client in figure 9.24.

1 package se.leiflindback.oodbook.polyminherit.tempmet;
2
3 /**
4 * The abstract superclass providing the template method.
5 */
6 public abstract class TaskTemplate {
7 public void performTask() {
8 // Some code, which is common for all concrete
9 // subclasses.
10 doPerformTask();
11 // Some code, which is common for all concrete
12 // subclasses.
13 }
14
15 /**
16 * Subclasses must provide an implementation of
17 * this method.
18 */
19 protected abstract void doPerformTask();
20 }

Listing C.217 Java code implementing TaskTemplate in figure 9.24.

1 package se.leiflindback.oodbook.polyminherit.tempmet;
2
3 /**
4 * A concrete implementation of the template.
5 */
6 public class ConcreteTaskA extends TaskTemplate {
7 @Override
8 protected void doPerformTask() {
9 // Some code, which is specific for ConcreteTaskA.
10 }
11 }

Listing C.218 Java code implementing ConcreteTaskA in figure 9.24.

1 package se.leiflindback.oodbook.polyminherit.tempmet;

344
 Appendix C Implementations of UML Diagrams

2
3 /**
4 * A concrete implementation of the template.
5 */
6 public class ConcreteTaskB extends TaskTemplate {
7 @Override
8 protected void doPerformTask() {
9 // Some code, which is specific for ConcreteTaskB.
10 }
11 }

Listing C.219 Java code implementing ConcreteTaskB in figure 9.24.

345
Bibliography
[LAR] C. Larman: Applying UML and Patterns, third edition, Prentice-Hall 2004,
ISBN:0131489062

[CW] Course web for Object Oriented Design, IV1350 http://www.kth.

se/social/course/IV1350/page/lecture-notes-60/

[JCC] The original Java code convention http://www.oracle.

com/technetwork/java/javase/documentation/
codeconvtoc-136057.html

[JCS] Java Coding Standars from Ambysoft Inc. http://www.ambysoft.

com/downloads/javaCodingStandardsSummary.pdf

[FOW] M. Fowler: Refactoring: Improving the Design of Existing Code, Addison-

Wesley 1999, ISBN: 9780201485677

[JU] Home page for the JUnit unit testing framework http://junit.org/
junit4/

[NB] Home page for the NetBeans IDE http://netbeans.org/

[Code] A Git repository containing a NetBeans project with all the source code in
this text, https://github.com/oodbook/code The complete project
can also be downloaded as a zip file, https://github.com/oodbook/
code/archive/master.zip

[GOF] E. Gamma, R. Heml, R. Johnson, J. Vlissides: Design Patterns: Ele-

ments of Reusable Object-Oriented Software, Addison-Wesley 1994, ISBN:
0201633612

[Java Tutorial] https://docs.oracle.com/javase/tutorial/

java/IandI/defaultmethods.html, accessed Feb 22, 2017

346
Index

@Override, 10 composite, 214

composition, 183
activation bar, 22, 40 constant, 102
actor, 34 constraint, 149
alternative flow, 18 constructor, 3, 41, 42
analyses, 15, 20 controller, 52
anchor, 24 coupling, 48, 180
annotation, 9, 123
architectural pattern, 50 design, 15, 39
architecture, 50 concept, 43
array, 5 method, 57
assert, 123 mistake, 78
association, 21, 29 design pattern, 192
direction of, 21 composite, 214
multiplicity, 22, 29 factory, 206
name, 21 observer, 192
attribute, 21, 28 singleton, 211
strategy, 199
basic flow, 17
template method, 218
business logic, 51
dictionary, 225
category list, 24 domain model, 24
checked exception, 154 naïve, 31
class, 2, 21, 24 programmatic, 31
class candidate, 24 DTO, 55
class diagram, 20 duplicated code, 84
code convention, 81
encapsulation, 43, 179, 180, 184
code smell, 83
exception, 6, 149
coding, 15, 58
checked, 7
mistake, 118
runtime, 7
cohesion, 46, 180
collection, 5 factory, 206
combined fragment, 23 found message, 40
comment, 24 framework, 122
comments, 161
communication diagram, 42 Gang of Four, 192

347
 Index

guard, 37 package, 50
package diagram, 40
high cohesion, 46 package private, 50
immutable, 67, 163 pattern, 50, 51, 54, 55
implementation, 43 polymorphism, 175
inheritance, 10, 156, 182 primitive variables, excessive use, 96
integration, 15 programming, see coding
interaction diagram, 42, 58, 61 protected, 182
interaction operand, 23 public interface, 43
interaction use, 41 refactoring, 83
intercation operator, 23 reference, 4
interface, 9 requirements analyses, 14
iteration, 13
sequence diagram, 22, 40
javadoc, 8, 161 singleton, 211
@param, 8 spider-in-the-web, 33, 48, 69, 78
@return, 8 state, 51, 163
JUnit, 121, 123 static, 3, 40, 41
large class, 90 stereotype, 41, 150, 174
layer, 54 strategy, 199
lifeline, 22 subclass, 10
list, 5 super, 11
long method, 88 superclass, 10
long parameter list, 91 SUT, 121
low coupling, 48 system operation, 34, 52, 57, 61
system sequence diagram, 34
meaningless name, 101 system under test, 121
member, 40
message, 22 template method, 218
methodologies, 13 test, 14, 15, 120
model, 51 this, 3
MVC, 51 type, 12

naming convention, 23 UML, 15, 20, 39

naming identifier, 101 unchecked exception, 154
NetBeans, 129 unit test, 121
new, 4 unnamed values, 102
note, 24 utility class, 80
noun identification, 24 view, 51
object, 2, 22 visibility, 40, 43, 50, 182
observer, 192
operation, 21
overload, 11

348