Java Language Definition

This section defines the Java subset used in this text. (Consult The Java Language Specification, by J. Gosling, B. Joy, and G. Steele, Addison Wesley Publishing, 1996, for the definition of the full Java language.)

A language is defined by its syntax (its spelling and grammar) and its semantics (what its sentences mean). We will define syntax with the grammatical notation explained in the next section, and we will give semantics by means of English explanation and examples. Also, references will appear to the relevant sections of the text.

Grammatical Notation

The syntax of a computer language are defined by a set of grammar rules or ``equations'' that define the phrases' precise appearance. An example best introduces this format.

Using an English description, we might define a number to appear as a whole-number part followed by an optional fraction part. Both whole-number and fraction parts consist of nonempty sequences of numerals; the fraction part is preceded by a decimal point. These grammar rules formalize the description:

NUMERAL ::=  0 | 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9
The words in upper-case letters name the phrase forms. Thus, the first rule tells us that a NUMBER phrase consists of a WHOLE_PART followed by a FRACTION_PART; the question mark states that the FRACTION_PART is optional. The second and third rules have obvious readings; note the decimal point in the third rule.

The asterisk in the fourth rule stands for ``zero or more''---a NUMERAL_SEQUENCE consists of one NUMERAL followed by zero or more additional NUMERALs. The vertical bars in the last rule are read as ``or.'' Unless stated otherwise, appearances of ?, *, and | in grammar rules will stand for the notions just described.

For example, 123.4 is a NUMBER because 123, the WHOLE_PART, is a NUMERAL_SEQUENCE and because .4 is a FRACTION_PART, where 4 is a NUMERAL_SEQUENCE.

For conciseness, the first three grammar rules can be compressed into just one:

The double-bracket pairs, [[ and ]], enclose a phrase; because of the question mark suffixed to it, the entire phrase is optional. We also use double brackets to enclose phrases that can be repeated zero or more times, e.g.,
This rule defines the syntax of a list of numbers separated by commas, e.g., 1, 2.34, 56.7 is a NUMBER_LIST as is merely 89.

Double brackets can be used to enclose alternatives, e.g., this grammar rule defines exactly the two words, ``cat'' and ``cut'':

WORDS ::=  c [[ a | u ]] t

As the examples show, spaces within the grammar rules do not imply that spaces are required within the phrases defined by the rules. For example, we normally write a NUMBER, like 56.7, with no internal spaces. We follow this convention in the sections that follow.

Java Definition

We now define the Java subset used in this text. For exposition, we present the language one construct at a time. Each construct is defined by

Program Unit

PROGRAM_UNIT ::=  [[ package IDENTIFIER ; ]]?
                 [[ CLASS | INTERFACE ]]
PACKAGE_NAME ::=  IDENTIFIER  [[ . IDENTIFIER ]]* [[ . * ]]? ;
A Java program unit begins with an optional package affiliation, followed by zero or more import statements, followed by either a class definition or an interface definition. (Important: the asterisk in the phrase, [[ . * ]], is indeed an asterisk to be included in the package name; it is not indicating a repetition.) An IDENTIFIER is a name, as defined in ``What is an Identifier?'' in Chapter 3. CLASS and INTERFACE are defined below.


package MyGraphicsFrames;
import java.util.GregorianCalendar;
import javax.swing.*;
import java.awt.*;
import java.awt.event.*;
public abstract class MyFrame extends JFrame
{ ... }

Semantics: A package affiliation groups a Java unit with others in the same-named package (see ``Packages'' in Chapter 9). The import statements make available the classes contained in the imported packages (see Figure 4 and ``Java Packages'' in Chapter 2). When the . * suffix is omitted, then the PACKAGE_NAME must name a specific class in a specific package, and it is only this one class that is available for use.


CLASS ::=  public [[ abstract | final ]]?
           { FIELD*
As indicated by the grammar rule, the keywords, abstract and final are exclusive and optional. The identifier mentioned in EXTEND must be the name of a class; the identifiers in IMPLEMENT must be names of interfaces. FIELD, CONSTRUCTOR, and METHOD are defined in subsequent sections.


public abstract class MyFrame extends Frame implements MyInterface
{ private int i = 2;
  public MyFrame() { ... }
  public void f() { i = i + 1; }
  public void paint(Graphics g) { ... }
  public abstract int h(GregorianCalendar g);

Semantics: A class defines a ``blueprint'' from which objects are constructed. Chapter 2 introduces classes; see also ``Revised Syntax and Semantics of Classes'' at the end of Chapter 5.

An abstract class lacks codings for some of its methods, and objects cannot be constructed directly from an abstract class (see ``Abstract Classes'' in Chapter 9)---another class must be written that extends the abstract class. A final class cannot be extended (see ``final Components'' in Chapter 9). A class can extend at most one other class but it can implement multiple interfaces.


INTERFACE ::=  public interface IDENTIFIER EXTEND?
               {  [[ public abstract? RETURN_TYPE METHOD_HEADER ; ]]* }
An interface contains zero or more header lines of methods. RETURN_TYPE and METHOD_HEADER are defined below in the sections, ``Type'' and ``Constructor,'' respectively.


public interface MyInterface
{ public void f();
  public int h(GregorianCalendar g);

Semantics: An interface lists method names that a class might possess; a class that does so is said to implement the interface. See ``Interfaces'' in Chapter 9 for details and examples.


VISIBILITY ::=  public | private | protected | (nothing)
DECLARATION is defined in a later section.


private int i = 2;
public static final double ONE = 1.0;
private static boolean b;
private GregorianCalendar today = new GregorianCalendar();

Semantics: A field is a variable declaration that exists independently of the methods that use it. (See ``Objects with State: Field Variables'' and ``Scope of Field Declarations'' in Chapter 4.) A final field cannot not be altered in value once it is initialized. A static field can be used by static methods, like main. As illustrated in ``Case Study: Card Games,'' in Chapter 8, a public static final field defines a constant name that can be used throughout a program.

The forms of visibility are

Public and private visibilities are used extensively in examples throughout Chapter 5. Fields are almost always labelled private. Protected visibility is used occasionally with subclasses---see ``Subclasses and Method Overriding'' and Figure 15 in Chapter 9 for an example. The last visibility, called package visibility, is not used in this text.


                 { [[ super ( ARGUMENT_LIST ) ; ]]?
A constructor method's name must be the same as the class in which the constructor is included. A class can have more than one constructor method, provided that each constructor is uniquely identified by the quantity and data types of its formal parameters. (Such a situation is called overloading the constructor.) The formal parameters in the METHOD_HEADER must be distinctly named.

If the class in which the constructor appears extends another class, then the first statement of the constructor can be an invocation of the constructor in the super class---the invocation is called super. STATEMENT is defined in a later section.

ARGUMENT_LIST is defined in the section, ``Invocation.'' TYPE is defined below.


public MyFrame()
{ setSize(200, 100);

public MyFrame(int width, String title)
{ setTitle(title);
  setSize(width, width/2);

public MyCustomButton(String label)
{ super(label); 

Semantics: When an object is constructed from a class, the object is constructed in computer storage and the class's constructor method is executed. Chapter 5 introduces and uses extensively constructor methods; see ``Revised Syntax and Semantics of Classes'' at the end of that chapter for technical details. The semantics of executing a constructor method is the same as the semantics of executing a method, and the section on methods that follows should be studied.


PRIMITIVE_TYPE ::=  boolean | byte | int | long | char | float | double
RETURN_TYPE ::=  TYPE | void
Note that type names like String, GregorianCalendar, etc., are in fact identifiers.



Semantics: As explained in ``Data Type Checking'' in Chapter 3, types in Java are classified as primitive or reference types. Elements of the former are atomic, non-object values; the latter are objects. Every class and interface defines a data type; objects constructed from a class are the elements of that class's type. If a class implements an interface, the class's objects also belong to the interface's type.

Types are related by a subtyping relation; this is explained in Appendix II. A RETURN_TYPE is the type of value that a method may return as its result; void means ``no result returned.''


METHOD ::=  VISIBILITY  [[ abstract | static ]]?  final?  RETURN_TYPE
The method is labelled abstract if and only if the method's body consists of just a semicolon. A static method is invoked by other static methods, e.g., main. A final method cannot be overridden (see below). A method's formal parameters must have distinct names.


private double sum(int i, double j)
{ return i + j; }

public abstract int h(GregorianCalendar g);

public static void main(String[] args)
{ new MyFrame(300, "My Test Frame"); } 

Semantics: A method is invoked by a client, which supplies arguments to the formal parameters in the method's header. The arguments bind to the formal parameter names, and the statements in the method's body execute. See Chapter 5 for extensive examples and see ``Formal Description of Methods'' in Chapter 5 for a precise description of invocation, binding, and execution.

A class can have multiple methods with the same name, provided that each method is uniquely identified by the quantity and data types of its formal parameters. (Such a situation is called overloading the method name; see ``Method Overloading'' and ``Semantics of Overloading'' in Chapter 9 for details.)

If a class extends a superclass, then the class can contain a method whose name and formal parameters are identical to those of a method in the superclass. This is called overriding the method. See ``Subclasses and Method Overriding'' and ``Semantics of Overriding'' in Chapter 9 for extensive explanation.


              | RETURN
              | IF | SWITCH
              | WHILE | DOWHILE | FOR
              | TRY_CATCH
              | THROW
              | STATEMENT_EXPRESSION ;
              | { STATEMENT* }
              | ;
The various statement forms are defined in the sections that immediately follow.

Examples and Semantics: Statements are executed in sequence, from ``left to right.'' Statement expressions are statements that return results; a statement expression can be used where an expression is expected. Examples are

GregorianCalendar g = new GregorianCalendar();  // object construction
System.out.println(Math.abs(-2) + 3);  // invocation
int i;  int j = (i = 3) + 1;  // assignment: set i to 3 and then j to 4
The last usage of a statement expression is not encouraged.


INITIAL_EXPRESSION is defined in ``Expression''; its data type must be a subtype of the TYPE in the declaration.


int i = 2 + 3;
GregorianCalendar[] days_in_this_month;

Semantics: A declaration creates a cell, named by the identifier, to hold values of the indicated data type. The cell receives the value computed from the initial expression (if any). The sections, ``Declarations and Variables'' ``Local Variables and Scope,'' in Chapter 3 provide examples and details.


The statement can appear only in the body of a method that has a non-void return type stated in its header line; the data type of EXPRESSION must be a subtype of the return type. (See Section ``Type'' for the definition of RETURN_TYPE.)


return 2 * sum(3, 4);
return new int[3];
return i;

Semantics: The return-statement calculates a value, which is immediately returned as the result of a method. See ``Results from Methods'' in Chapter 5.


IF ::=  if ( EXPRESSION ) { STATEMENT* } [[ else { STATEMENT* } ]]?
SWITCH ::=  switch ( EXPRESSION )
            { [[ case EXPRESSION : { STATEMENT*  break ; } ]]*
              default : { STATEMENT* }
The expression part of the if-statement must have data type boolean. The first expression of the switch-statement must have data type integer, char, or byte.

The second expression of the switch-statement must be an integer-, char-, or byte-typed constant expression, that is, an integer expression that is constructed from numeric constants, arithmetic operations, and fields that are static and final. Here is an example that uses a convoluted numeric constant:

public class Test
{ private static final int k = 2;

 public static void main(String[] args)
  { int x = 0;
    switch (x + 1)
           { case ((int)(k / 3.3) + 1) : { break; }
             default : { } 
The Java compiler calculates the integer represented by the constant expression and replaces the expression by the integer. (Here, (int)(k / 3.3) + 1 is replaced by 1.) In practice, constant expressions are merely integer and character constants, e.g., 5 and 'a'.

Additional Example:

if ( i > 0  &&  i != 2 )
     { j = sum(i, 2.2); }
else { if ( b )
          { j = 0; }

Semantics: The appropriate arm of an if- or switch-statement is selected based upon the value of the test expression. Chapter 6 presents numerous examples of if-statements; see ``The Switch Statement'' at the end of that chapter for examples of the switch-statement.


DOWHILE ::=  do { STATEMENT* } while ( EXPRESSION ) ;
FOR ::=  for ( [[ DECLARATION | ASSIGNMENT ; ]] 
               EXPRESSION ;
             { STATEMENT* }
The test expression for each of the three loops must be typed boolean. Although the for-loop accepts a STATEMENT_EXPRESSION as the third part of its header line, in practice this is an assignment statement. Alas, there is no restriction regarding the variables declared, referenced, and assigned in the three parts of the for-loop's header line.


for ( int i = 0;  i != r.length;  i = i + 1 )
    { while ( r[i] > 0 )
            { r[i] = r[i] - 2; }

Semantics: As long as the test expression evaluates to true, the body of a loop repeats. Chapter 7 displays numerous examples. The do-while-loop is a minor variation of the while-loop---the body executes once before the loop's test is computed; see ``The do-while loop'' in Chapter 7.

The for-loop executes its declaration-assignment on entry; next, the test expression is computed, and as long as the text computes to true, the loop's body is executed, followed by the statement-expression in the loop's header line. For loops are introduced in ``The for-loop'' in Chapter 7 and are extensively used with arrays in Chapter 8.

Exception Handler

The data type listed in the CATCH clause of an exception handler and the data type of the EXPRESSION in the THROW statement must be a subtype of Throwable.


try { if ( i != 0 )
           { j = j / i; }
      else { throw new RuntimeException("i is zero"); }
catch (RuntimeException e)
      { System.out.println(e.toString()); 
        j = 0;

Semantics: The statements in the try clause of an exception handler are executed as usual, unless a throw statement is executed. The throw aborts usual execution and forces an immediate search of the clauses in the CATCH_LIST for the first catch clause whose data type is a subtype of the type of the exception object listed in the throw statement. If a matching catch clause is found, the exception binds to its formal parameter, the clause's statements are executed. If no matching catch clause is found, search for a matching clause proceeds at the next enclosing exception handler, as found by search the prior sequence of statements that were executed. See ``Exceptions Are Objects'' in Chapter 11 for details.

Object Construction

                        | new ARRAY_ELEMENT_TYPE DIMENSIONS
DIMENSIONS ::=  [ EXPRESSION ]  [[  [ EXPRESSION ]  ]]* [[  []  ]]*
Objects are constructed in two forms: individual objects and array objects. For constructing the former, the identifier must be a class name, and its argument list must contain the same quantity of parameters as (one of) the constructor methods included in the class; the data type of each argument must be a subtype of the data type listed with the corresponding formal parameter in the constructor method.


new GregorianCalendar()
new MyFrame(300, "")

When array objects are constructed, the data type of the array's individual elements is listed first, followed by the dimensions. The individual elements can be of primitive type or a type defined by a class name. The dimensions are indicated by pairs of brackets, where the size of each dimension is indicated by an expression embedded within a bracket pair. The quantity of at least the first dimension must be indicated.


new int[4]
new GregorianCalendar[12][]
new double[size][size * size][][]

Semantics: The statement creates an object from class IDENTIFIER (or an array that holds values of the array-element type). Examples pervade the text---see ``An Application that Creates an Object'' in Chapter 2 for a simple one, see ``Revised Syntax and Semantics of Classes'' in Chapter 5 for details about object creation, and see ``Formal Description of Arrays'' at the end of Chapter 8 for examples of array object construction. The class's constructor method whose formal parameters best match the argument list is executed; see ``Formal Description of Methods'' in Chapter 5 for details about the matching process.

Method Invocation

RECEIVER ::=  this | super
            | IDENTIFIER
            | RECEIVER [ EXPRESSION ]
            | ( RECEIVER )
The identifier mentioned in the INVOCATION rule is a method name; the identifier in RECEIVER can be either a variable name whose data type is a reference type (that is, the variable holds the address of an object) or it can be the name of a class, which occurs when a static method is invoked; see below.

The argument list must contain the same quantity of parameters as the number of formal parameters listed in the header line of the invoked method, IDENTIFIER. The data type of each argument must be a subtype of the data type listed with the corresponding formal parameter.

Examples and Semantics: To send a message (that is, to invoke a method), the (address of the) object where the method lives must be determined---the object is the ``receiver'' of the message. If the RECEIVER is omitted from the invocation, then the receiver is the same object where the invocation is situated. Private methods are usually invoked this way:

public class C
{ private static double sum(int i, double j)
  { return i + j; }

  public static void main(String[] args)
  { System.out.println( sum(2, 3.4) ); }
Public methods can be invoked this way, also:
import javax.swing.*;
public class MyFrame extends JFrame
{ public MyFrame()
  { setTitle("MyFrame");

  public void f(String label)
  { this.setTitle(label); } 

All invocations in the example ultimately invoke setTitle in the superclass, JFrame; see ``Customizing Frames with Inheritance'' in Chapter 4. The semantics of this is the same as an omitted RECEIVER.

super asserts that the receiver should be the same object as the invocation's but the invoked method must be selected from the superclass from which the object is built:

import javax.swing.*;
public class MyFrame extends JFrame
{ public MyFrame() 
  { super.setTitle("MyFrame"); }

  public setTitle(String s)
  { }
invokes the setTitle method located in class JFrame and ignores the one in MyFrame. See Figure 16, Chapter 9, for a significant example.

When an identifier defines the receiver, the identifier is usually a variable that holds the address of an object, as g does here:

GregorianCalendar g = new GregorianCalendar();
... g.getDate(); ...
But the identifier can also be a class name when the invoked method is static. Here is an example:
public class C
{ public static int one() { return 1; } }

public class D
{ public static void main(String[] args)
  { System.out.println( 2 * ); }
Table 9 of Chapter 3 lists a variety of static methods that are invoked with the receiver, Math, e.g., Math.sqrt(2).

A receiver can be indexed if it names an array object, e.g.,

GregorianCalendar[] dates = new GregorianCalendar[3];
System.out.println( dates[0].getTime() );
Similarly, an object that contains a public field that holds an object can be indexed by an identifier:
public class C
{ public GregorianCalendar g = new GregorianCalendar();

public class D
{  ...
  C x = new C();
  ...  x.g.getTime() ...

Finally, a statement expression can define a receiver. This happens when an object is constructed anew, e.g., new GregorianCalendar().getTime(), and when a method invocation returns an address of an object as its result:

public class C
{ public GregorianCalendar f() { return new GregorianCalendar(); }

public class D
{ ... (new C().f()).getTime() ...  }
Here, new C().f() is the invocation that defines the receiver for the invocation of getTime.

Once the receiver is established, the method within the receiver is selected. (Note the impact of super on this selection.) Next, the arguments are evaluated, bound to the formal parameters of the selected method, and the selected method executes. Precise descriptions of these steps are found in ``Formal Description of Methods'' in Chapter 5, ``Semantics of Overloading'' in Chapter 9, and ``Semantics of Overriding'' also in Chapter 9.

Examples of method invocation abound in the text; Chapter 5 is devoted to this one topic.


              | RECEIVER [ EXPRESSION ]
              | RECEIVER . IDENTIFIER
The data type of the assignment's expression must be a subtype of the variable's data type.

Examples and Semantics: An assignment places the value of its right-hand side expression into the cell named by the left-hand side variable. Variables are usually identifiers, e.g., x in int x; x = 0.

A variable can be the element of an array, as is r[2] in int[] r = new int[3]; r[2] = 0. In this case, the address of the array object, r is first computed as a RECEIVER, and the address is used with the index. (See the explanation of RECEIVER in the previous section.)

If an object possesses public fields, the variable part of an assignment can be a reference to one of those fields:

public class C
{ public int x; }

public class D
{ ...  C c = new C();
       c.x = 1;  ...


                | VARIABLE
                | ( EXPRESSION )
                | ( TYPE ) EXPRESSION
                | EXPRESSION instanceof REFERENCE_TYPE
                | this | null
                | STATEMENT_EXPRESSION
INFIX_OPERATOR ::=   +  |  -  |  *  |  /  |  % 
                     |  <  |  >  |  <=  |  >=  |  ==  |  !=
                     |  ||  |  &&
PREFIX_OPERATOR ::=  -  |  !
                        | { [[ INITIAL_EXPRESSION_LIST ]]? }
As always, infix and prefix operators require arguments whose data types are acceptable to the operators, e.g., the multiplication operator, *, requires arguments of numeric data type. Literals are constants, e.g., 3 and true; see Table 2 of Chapter 3.

INITIAL_EXPRESSIONs are used to initialize variable declarations; the set expressions are used to initialize array variables, e.g., int[][] r = { {0,1}, {1,3,5} {}, {1}}, constructs and initializes an array with 4 rows where each row has a different number of columns.

Semantics and Examples: As described in Chapter 3, expressions are computed from left to right, producing a result that can be a primitive value or the address of an object. Consider each of the clauses of the syntax definition for EXPRESSION:

Literal and Identifier

The forms of literals are listed in Table 2, Chapter 3. Identifiers are defined by the section, ``What is an Identifier?,'' at the end of Chapter 3.

Typing and Subtyping

There are two forms of data types in the Java language:

Every class and interface definition defines a reference data type, e.g., public class MyFrame extends JFrame defines type MyFrame, and public interface UpdateableView defines type UpdateableView.

The Java compiler calculates data types to enforce compatibility in expression calculation, assignment, parameter binding, and results returned from methods. For example, given

int i = 200;
JFrame f = new JFrame();
f.setSize(i + 1, i / 2); 
the compiler calculates data types of the numerals, variables, and expressions to validate in advance of these statements' execution that they will behave correctly: Because of the compiler's work, none of the above checks must be repeated when the program executes.


The Java compiler allows some flexibility in its data type calculation---given a context where a value of data type, T1, is required, the Java compiler will allow a value whose data type, T2, is a subtype of T1. We write T2 <= T1. One simple example is

int i = 3;
double d = i * 2.5;
Although i is int-typed, it is allowed in the context, double d = __ * 2.5, because int <= double, that is, int is a subtype of double.

Another context is the binding of actual parameters to formal parameters: A method that expects an argument of type T1 will execute correctly if it receives an argument of type T2 such that T2 <= T1 holds true. An example is

private void f(double x) { ... }
Within the context, f(__), a double is required, but 3 suffices because its data type is int and int <= double.

Subtypes prove crucial to object usage, because data types are embedded into objects when objects are constructed during program execution. Consider

Component f = new JFrame();
if ( f instanceof JFrame )
   { ((JFrame)f).setVisible(true); }
the JFrame object constructed in the first statement has an internal data type of JFrame, even as the Java compiler has calculated that f holds values of type Component. The assignment is approved by the compiler because JFrame <= Component; see Chapter 10.

When the Java compiler calculates data types, it does not execute the statements themselves, and when the compiler examines the above if-statement, it knows merely that f has data type Component; it cannot answer whether the object held by f is truly a JFrame.

Later, when the program is started and the if-statement executes, the instanceof operation examines the data type embedded within the object named by f to calculate the answer to the test. If the answer is true, the cast operation lets f be used as a JFrame-typed object.

Subtypes of Primitive Types

The primitive types used in this text are boolean, char, byte, int, long, float, and double; see Chapter 3. The numeric types in the above list are related by subtyping as follows:
byte  <=  int  <=  long  <=  float  <=  double
where T1 <= T2 if T1 lies no further to the right in the ordering than does T2. For example, byte <= int, byte <= long, and int <= int.

Subtypes of Reference Types

Reference types are related by subtyping according to the following rules. Given arbitrary reference types, T1 and T2, we say that T1 <= T2 when one of the following rules holds true:

For example, from these definitions,

public interface I1 { ... }

public interface I2 extends I1 { ... }

public class C implements I1 { ... }

public class D extends C implements I2 { ... }
we can conclude, say, D[] <= I1[], because Yet a second way of concluding the same fact is by noting that class D implements I2, interface I2 extends I1, hence, D <= I1 and then D[] <= I1[].

Class Diagrams and Hierarchies

Programs are designed and documented with class diagrams. A class diagram consists of a collection of components (names of classes and interfaces) connected by lines and arrows that represent collaborations and dependencies. The class diagrams used in this text are drawn in a subset of the diagram language of UML (Universal Modelling Language; see UML Distilled, by M. Fowler, Addison-Wesley, 1999, for a concise introduction.) Here is a summary of the diagram language.


In its simplest representation, a class is drawn as a rectangle containing the class's name.

A class can be presented with some or all of its public methods:

A method can be specified with just its name, e.g., deposit, or it can be listed with its formal parameters and its result type, e.g., withdraw(int amount): boolean. Note that the result type is listed at the right of the method's specification, preceded by a colon. Usually, constructor methods are not listed with the class, but they can appear if the constructor or its parameters are important to understanding the class.

A class's attributes can be listed as well:

When an attribute is specified, one can assume that the class will have an accessor method for the attribute, e.g., public int balanceOf() { return balance; }. See Figure 12, Chapter 6, for this example.

Objects can be that already exist prior to the execution of a program can be included in a class diagram as well:

An abstract class is specified by writing in italics the class's name and those methods that are abstract. (If you write the class diagram by hand, write the word, ``abstract'' in front of those words that should be in italics.)

See Figure 18, Chapter 9, for this example.

An interface is drawn like an abstract class---the name and all methods appear in italics:

Such an arrangement might appear at an early stage of design, before specifics about methods and dependencies are decided.

An arrow from one component to another means that the first component depends on the second. By ``depends,'' we mean some or all of the following:

Here is an example from Chapter 7:

A stronger form of dependence is called composition, which defines the ``has-a'' relationship:

The large, black arrowheads in the diagram assert that every JPanel object will have its own unique Graphics object, and every ClockWriter2 will have its own GregorianCalendar.

When one class extends another by means of inheritance, an arrow with a head is drawn from the subclass to the superclass:

The arrow indicates that an object created from the subclass will contain the structure inherited from the superclass; in particular, the subclass's methods include methods listed in the superclass. Here, ClockWriter2 is a JPanel extended by additional methods.

When a component depends on (refers to) an interface, a dotted arrow is drawn from the component to the interface:

Here, the coding of MortgageCalculator depends on interface BankAccountSpecification; see Chapter 9 for this example.

A dotted arrow with an arrow head is drawn from a class to an interface when the former implements the latter:

Here, class BankAccount implements BankAccountSpecificaiton.


The usual dependency between components is one-to-one, that is, if an arrow is drawn from component A to component B, it suggests that every one A object depends on one B object. To make explicit other quantities of dependency, we annotate the arrow with integers. For example, Figure 17 of Chapter 6 displayed the class diagram of an application whose controller manage two bank accounts and their output views:

The annotations on the arrow state that each one AccountController object depends upon (uses) two distinct BankAccount objects.

If there is a one-to-many dependency, but the exact quantity of the ``many'' is determined only when the application is built, an asterisk can be used in place of the integer. For example, the slide puzzle application in Chapter 8 is designed to construct slide puzzles with a varying quantity of movable puzzle pieces. Therefore, the puzzle board uses at least one and possibly many puzzle pieces:

Class Hierarchies

A package of classes that are related by inheritance is depicted in a form of class diagram that uses indentation in place of large arrow heads. For example, if class Window extends Container, JFrame extends Window, and JApplet extends Container, we might display the classes in this hierarchy:
 |  |
 |  +-JFrame
Figure 5, Chapter 10, and the API Web pages for the Java packages display such class hierarchies.