Spring Semester, 1999
Coupling
© 1999, All Rights Reserved, SDSU & Roger Whitney
San Diego State University -- This page last updated 28-Jan-99
|
|
CS 535 Object-Oriented Programming & Design Spring Semester, 1999 Coupling |
|
|---|---|---|
|
© 1999, All Rights Reserved, SDSU & Roger Whitney San Diego State University -- This page last updated 28-Jan-99 |
Quality of ObjectsNot all objects are created Equal!
Metrics for quality
Coupling
Coupling
Decomposable system
Coupling
Measure of the interdependence among modules
"Unnecessary object coupling needlessly decreases the reusability of the coupled objects "
"Unnecessary object coupling also increases the chances of system corruption when changes are made to one or more of the coupled objects"
Types of Modular CouplingIn order of desirability
Data Coupling (weakest – most desirable)
Control Coupling
Global Data Coupling
Internal Data Coupling (strongest – least desirable)
Content Coupling (Unrated)
Modular Coupling Data Coupling
Output from one module is the input to another
Using parameter lists to pass items between routines
Common Object Occurrence:
class Receiver
{
public void message( MyType X )
{
// code goes here
X.doSomethingForMe( Object data );
// more code
}
}
Modular
Coupling
Data Coupling
Major
Problem
Example: Sorting student records, by ID, by Name
How does the SortedList method add() add the new student record object and resort the list? To do this it needs to access the ID (name) fields of StudentRecord!
class StudentRecord { Name lastName; Name firstName; long ID; public Name getLastName() { return lastName; } // etc. } SortedList cs696 = new SortedList(); StudentRecord newStudent; //etc. cs696.add ( newStudent );
Solution 1 Bad News
Here the add method actually accesses the StudentRecord method to get the ID. What is wrong with that? Why is this bad news?
class SortedList { Object[] sortedElements = new Object[ properSize ]; public void add( StudentRecord X ) { // coded not shown Name a = X.getLastName(); Name b = sortedElements[ K ].getLastName(); if ( a.lessThan( b ) ) // do something else // do something else } }Solution 2 Send message to object to compare self to another StudentRecord Object
How is this any better that solution 1? Is it any better? How does it differ?
class SortedList{ Object[] sortedElements = new Object[ properSize ]; public void add( StudentRecord X ) { // coded not shown if ( X.lessthan( sortedElements[ K ] ) ) // do something else // do something else } } class StudentRecord{ private Name lastName; private long ID; public boolean lessThan( Object compareMe ) { return lastName.lessThan( compareMe.lastName ); } etc. }Solution 3 Interfaces or "required operations"
Notice how the SortedList is no longer coupled to the StudentRecord class. It can be used to sort any list of objects of the same class than implement Comparable.
interface Comparable { public boolean lessThan( Object compareMe ); public boolean greaterThan( Object compareMe ); public boolean equal( Object compareMe ); } class StudentRecord implements Comparable { private Name lastName; private long ID; public boolean lessThan( Object compareMe ) { return lastName.lessThan( ((Name)compareMe).lastName ); } } class SortedList { Object[] sortedElements = new Object[ properSize ]; public void add( Comparable X ) { // coded not shown if ( X.lessthan( sortedElements[ K ] ) // do something else // do something else } }Solution 4 Function Pointers
Code is neither legal C/C++ nor Java. The idea is to pass in a function pointer to the SortList object, which it uses to compare the objects in the list.
typedef int (*compareFun ) ( StudentRecord, StudentRecord ); class SortedList { StudentRecord[] sortedElements = new StudentRecord[ properSize ]; int (*compare ) ( StudentRecord, StudentRecord ); public setCompare( compairFun newCompare ) { compare = newCompare; } public void add( StudentRecord X ) { // coded not shown if ( compare( X, sortedElements[ K ] ) ) // code not shown } } int compareID( StudentRecord a, StudentRecord b ) { // code not shown } int compareName( StudentRecord a, StudentRecord b ) { // code not shown } SortedList myList = new SortedList(); myList.setCompair( compareID );
Functor Pattern
Functions as Objects
Functors are functions that behave like objects
They serve the role of a function, but can be created, passed as parameters, and manipulated like objects
A functor is a class with a single member function
Note 1: Functors violate the idea that a class is an abstraction with operations and state. Beginners should avoid using the Functor pattern, as they can lead to bad habits. The functor pattern is used here only as a last resort.
Note 2: The Command pattern is similar to the Functor pattern, but contains operations and state.
Note 3: For more information about patterns see: http://www.eli.sdsu.edu/courses/spring98/cs635/notes/patternIntro/patternIntro.html .For more information about Functor and Command pattern see: http://www.eli.sdsu.edu/courses/spring98/cs635/notes/command/command.html
Function Pointers in JavaComparator in Java 2 (JDK 1.2) In Java 2, the Comparator interface defines an interface for objects that act like functions pointers to compare objects.
Methods in Comparator Interface int compare(Object o1, Object o2)
Comparator Example import java.util. Comparator; class Student { String name; int id; public Student( String newName, int id ) { name = newName; this.id = id; } public String toString() { return name + ":" + id; } } final class StudentNameComparator implements Comparator { public int compare( Object leftOp, Object rightOp ) { String leftName = ((Student) leftOp).name; String rightName = ((Student) rightOp).name; return leftName.compareTo( rightName ); } public boolean equals( Object comparator ) { return comparator instanceof StudentNameComparator; } }
//Comparator Example Continued final class StudentIdComparator implements Comparator { static final int LESS_THAN = -1; static final int GREATER_THAN = 1; static final int EQUAL = 0; public int compare( Object leftOp, Object rightOp ) { long leftId = ((Student) leftOp).id; long rightId = ((Student) rightOp).id; if ( leftId < rightId ) return LESS_THAN; else if ( leftId > rightId ) return GREATER_THAN; else return EQUAL; } public boolean equals( Object comparator ) { return comparator instanceof StudentIdComparator; } }
//Comparator Example Continued import java.util.*; public class Test { public static void main(String args[]) { Student[] cs596 = { new Student( "Li", 1 ), new Student( "Swen", 2 ), new Student( "Chan", 3 ) }; //Sort the array Arrays.sort( cs596, new StudentNameComparator() ); for ( int k = 0; k < cs596.length; k++ ) System.out.print( cs596[k].toString() + ", " ); System.out.println( ); List cs596List = new ArrayList( ); cs596List.add( new Student( "Li", 1 ) ); cs596List.add( new Student( "Swen", 2 ) ); cs596List.add( new Student( "Chan", 3 ) ); System.out.println( "Unsorted list " + cs596List ); //Sort the list Collections.sort( cs596List, new StudentNameComparator() ); System.out.println( "Sorted list " + cs596List ); //TreeSets are aways sorted TreeSet cs596Set = new TreeSet( new StudentNameComparator() ); cs596Set.add( new Student( "Li", 1 ) ); cs596Set.add( new Student( "Swen", 2 ) ); cs596Set.add( new Student( "Chan", 3 ) ); System.out.println( "Sorted Set " + cs596Set ); } }
//Comparator Example Continued
OutputChan:3, Li:1, Swen:2, Unsorted list [Li:1, Swen:2, Chan:3] Sorted list [Chan:3, Li:1, Swen:2] Sorted Set [Chan:3, Li:1, Swen:2]
Sorting With Different Keys import java.util.*; public class MultipleSorts { public static void main(String args[]) { List cs596List = new ArrayList( ); cs596List.add( new Student( "Li", 1 ) ); cs596List.add( new Student( "Swen", 2 ) ); cs596List.add( new Student( "Chan", 3 ) ); Collections.sort( cs596List, new StudentNameComparator() ); System.out.println( "Name Sorted list " + cs596List ); Collections.sort( cs596List, new StudentIdComparator() ); System.out.println( "Id Sorted list " + cs596List ); TreeSet cs596Set = new TreeSet( new StudentNameComparator()); cs596Set.addAll( cs596List ); System.out.println( "Name Sorted Set " + cs596Set ); TreeSet cs596IdSet = new TreeSet( new StudentIdComparator() ); cs596IdSet.addAll( cs596List ); System.out.println( "Id Sorted Set " + cs596IdSet ); } }Output Name Sorted list [Chan:1, Li:2, Swen:1] Id Sorted list [Chan:1, Swen:1, Li:2] Name Sorted Set [Chan:1, Li:2, Swen:1] Id Sorted Set [Chan:1, Li:2]Solution 5 Function Pointers using generic types
typedef int (*compairFun ) ( <type>, <type> );
Modular Coupling Control Coupling
Passing control flags between modules so that one module controls the sequencing of the processing steps in another module
Common Object Occurrences:
class Lamp {
public static final ON = 0;
public void setLamp( int setting ) {
if ( setting == ON )
//turn light on
else if ( setting == 1 )
// turn light off
else if ( setting == 2 )
// blink
}
}
Lamp reading = new Lamp();
reading.setLamp( Lamp.ON );
reading.setLamp)( 2 );
Cure:
Decompose
the operation into multiple primitive operations
class Lamp {
public void on() {//turn light on }
public void off() {//turn light off }
public void blink() {//blink }
}
Lamp reading = new Lamp();
reading.on();
reading.blink();
Control
Coupling
Common
Object Occurrences:
class Test {
public int printFile( File toPrint ) {
if ( toPrint is corrupted )
return CORRUPTFLAG;
blah blah blah
}
}
Test when = new Test();
int result = when.printFile( popQuiz );
if ( result == CORRUPTFLAG )
blah
else if ( result == -243 )
Cure:
Use exceptions
How
does this reduce coupling?
class Test {
public int printFile( File toPrint ) throws PrintExeception {
if ( toPrint is corrupted )
throws new PrintExeception();
blah blah blah
}
}
try {
Test when = new Test();
when.printFile( popQuiz );
}
catch ( PrintException printError ) {
do something
}
Modular
Coupling
Global
Data
Coupling
Two
or more modules share the same global data structures
Common
Object Occurrence
:
A
method in one object makes a specific reference to a specific external object
A
method in one object makes a specific reference to a specific external object,
and to one or more specific methods in the interface to that external object
A
component of an object-oriented system has a public interface which consists of
items whose values remain constant throughout execution, and whose underlying
structures/implementations are hidden
A
component of an object-oriented system has a public interface which consists of
items whose values remain constant throughout execution, and whose underlying
structures/implementations are
not
hidden
A
component of an object-oriented system has a public interface which consists of
items whose values
do
not
remain constant throughout execution, and whose underlying
structures/implementations are hidden
A
component of an object-oriented system has a public interface which consists of
items whose values
do
not
remain constant throughout execution, and whose underlying
structures/implementations are
not
hidden
Internal
Data
Coupling
One
module directly modifies local data of another module
Common
Object Occurrence:
Modular Coupling Lexical Content Coupling
Some or all of the contents of one module are included in the contents of another
Common Object Occurrence :
Object Coupling
Very little is written about object coupling. For more information see “Managing Class Coupling: Apply the Principles of Structured Design to Object-Oriented Programming,” UNIX Review, Vol. 2, No. 1, May/June 1989, pp. 34-40.
Coupling measures the strength of the physical relationships among the items that comprise an object
Cohesion measures the logical relationship among the items that comprise an object
Interface coupling is the coupling between an object and all objects external to it. Interface coupling is the most desirable form of object coupling. Internal coupling is coupling among the items that make up an object.
Object Coupling Interface Coupling
Interface coupling occurs when one object refers to another specific object, and the original object makes direct references to one or more items in the specific object's public interface
Includes module coupling already covered
Weakest form of object coupling, but has wide variation
Sub-topics
Object Abstraction Decoupling
Assumptions that one object makes about a category of other objects are isolated and used as parameters to instantiate the original object.
Example: List items
C++ templates and Ada’s generics are the constructs Berard is talking about. Making the LinkedListCell a template removes any type specific code from the LinkedListCell class. This helps insure that the class can hold any type.
C++ Example class LinkedListCell { int cellItem; LinkedListCell* next; // code can now use fact that cellItem is an int if ( cellItem == 5 ) print( "We Win" ); } template <class type> class LinkedListCell#2 { type cellItem; LinkedListCell* next; // code does not know the type, it is just a cell item, // it becomes an abstraction }
Java Example Java does not support templates. Instead it supports Object as a root type. Using an Object as a type in the LinkedListCell class has some of the decoupling that Ada generics or C++ templates achieve. However, it provides only one category of objects (all of them). This solution that Smalltalk (with no compile time type checking) also supports. The no compile time type checking solution is a common source of flame wars in the net. Java interfaces can be used to achieve decoupling in the same situations as Ada generics or C++ templates.
class LinkedListCellA { int cellItem; LinkedListCell* next; if ( cellItem == 5 ) print( "We Win" ); } class LinkedListCellB { Object cellItem; LinkedListCell* next; if ( cellItem.operation1() ) print( "We Win" ); }
Selector Decoupling
Example: Counter object
class Counter{ int count = 0; public void increment() { count++; } public void reset() { count = 0; } public void display() { code to display the counter in a slider bar }
Display of Counter
"display" couples the counter object to a particular output type
The counter class can not be used in other setting due to this coupling
Better Counter Class
class Counter{ int count = 0; public void increment() { count++; } public void reset() { count = 0; } public String toString() {return String.valueOf( count );} }
Primitive Methods
A primitive method is any method that cannot be implemented simply, efficiently, and reliably without knowledge of the underlying implementation of the object
Primitive methods are:
Selectors
Selectors are encapsulated operations which return state information about their encapsulated object and do not alter the state of their encapsulated object
Replacing
public void display() { code to display the counter }
with
public String toString() {return String.valueOf( count );}
is an example of Selector decoupling.
By replacing a composite method (display) with a primitive method the Counter class is decoupled from the display device
This makes the Counter class far more useful
It also moves the responsibility of displaying the counter elsewhere
Constructors
Operations that construct a new, or altered version of an object
Java and C++ both have language constructs called constructors. Berard has in mind a larger class of operations than those. Often static methods are used as constructors to create new objects.
Berard’s example illustrating constructor decoupling is extremely vague. The fromString method below does make it clear what type of parameter is needed to create a new calendar object. One point to learn from his discussion is the desirability to have well defined interface to creating objects from primitive objects.
class Calendar { public void getMonth( from where, or what) { blah } }
class Calendar { public static Calendar fromString( String date ) { blah} }
Primitive Objects
Primitive objects are objects that are both:
Composite Object
Object conceptually composed of two or more objects
Heterogeneous Composite Object
Object conceptually composed from objects which are not all conceptually the same
The date class below is composed of three items that are the same type: ints. However, these ints represent different conceptual entities.
class Date{ int year; int month; int day; }
Homogeneous Composite Object
Object conceptually composed from objects which are all conceptually the same
list of names - each item is a member of the same general category of object – a name
Berard’s homogeneous composite objects are basically container objects.
Iterator
Allows the user to visit all the nodes in a homogeneous composite object and to perform some user-supplied operation at each node
Both Java and C++ support iterators
Passive Iterator
Neither Java nor C++ support passive iterators. Smalltalk does support them. In a passive iterator, you pass a method or function to the composite object, and the object then applies the method to all elements in the object. Passive iterators in Smalltalk are very powerful. Passive iterators require very minimal code to use. They require efficient ways to deal with method/functions as parameters. Only one passive iterator can be active on an object at a time.
class List { Object[] listElements = new Object[ size ]; public void do( Function userOperation ) { for ( int k = 0; k < listElements.length(); k++ ) userOperation( listElements[ k ] ); } }
In Main
List grades = new List(); aFunction = ( item ){ print( item ) }; grades.do ( aFunction );
Active Iterator Java (Enumeration, Iterator (JDK1.2), ListIterator (JDK1.2), StringCharacterIterator) and C++ (in STL) use active iterators.
List grades = new List(); Iterator gradeList = grades.iterator(); while ( gradeList.hasNext() ){ listItem = gradeList.next(); print ( listItem ); }
Java Enumeration/Iterator
|
Methods |
||
Enumeration |
Iterator |
ListIterator |
|
hasMoreElements() |
hasNext() |
hasNext() |
|
nextElement() |
next() |
next() |
|
|
remove() |
remove() |
|
|
|
nextIndex() |
|
|
|
hasPrevious() |
|
|
|
previous() |
|
|
|
previousIndex() |
|
|
|
add() |
|
|
|
set() |
Iterators and Coupling
Using iterators reduces coupling by hiding the details of traversing through elements of a collection. If one used the non-iterator method of accessing the elements of collections, it becomes a lot of work to replace the use of one collection with another. One might want to replace an array with a binary search tree for better performance.
Array int[] list for (int k = 0; k < list.length; k ++ ) System.out.println( list[k] );Vector Vector list for (int k =0; k < list.size(); k++ ) Sytem.out.println( list.elementAt( k ) );Binary Search Tree BinarySeachTree list Node current = list.root(); Stack previous = new Stack(); Previous.push( current ); while (current != null ) { a lot of code here }
Inside Internal Object Coupling
Coupling between state and operations of an object
The big issue: Accessing state
Changing the structure of the state of an object requires changing all operations that access the state including operations in subclasses
Solution: Access state via access operations
C++ implementation
Accessing State
C++ Example
class Counter{ public: void increment(void); private: int value; void setValue(int newValue); int getValue(void); }; void Counter::increment(void) //Increase counter by one { setValue(getValue() + 1); }; void Counter::setValue(int newValue) { value = newValue; }; int Counter::getValue { return value; };
Outside Internal Coupling from Underneath
Coupling between a class and subclass involving private state and private operations
Major Issues:
Outside Internal Coupling from the Side
Class A accesses private state or private operations of class B
Class A and B are not related via inheritance
Main causes:
Type and Inheritance
Type is a set of values and a set of operations that act on those values
