The essence of objects
1 The essence of objects

The essence of objects

CS 3500, Fall 2016

The Object Model

  • Every system is made of smaller sub-systems or components

    • System working = components working with each other

  • A component

    • Is a logical grouping of objects

    • Has a well-defined purpose and functionality

    • Completes a task by using its own functionality, or delegating to other components

  • An object

    • Is the basic “functional unit” in the object model

    • Has a singular, well-defined purpose

    • Communicates with other objects by passing messages (i.e. calling methods)

The Object Model outside software

  • We think of the computer as “one unit”

  • Set of components working together

    • A well-defined purpose

      • RAM: temp memory

      • Disk drive: perm memory

      • Processor: process

    • A well-defined boundary

      • RAM doesn’t process

      • Disk doesn’t cool system

The Object Model outside software

  • Hospital has various departments

    • Each department has well-defined purpose

      • Orthopedics

      • Cardiac Unit

    • Each department has a well-defined boundary

      • Cardiac unit doesn’t treat fractures

  • To a patient, the hospital is one unit

How are objects defined?

  • Each object has a specific purpose

    • Maintain customer records

    • Validate a credit card

    • “Behave like a stack”

  • Each object is created so that it can be used by other objects in a specific way

    • Object A assumes object B can handle a particular task

    • Objects of the same “type” are from the same class

  • Class=data type, Object=actual thing

The object model

  • Each object has

    • State

    • Behavior

  • State of an object

    • Each object has some data associated with it

      • Members or variables of the object

    • Values of that data together decide the “state” of the object

  • Behavior of the object

    • Each object has capabilities to operate on its or other data

      • Operations are in the form of methods or functions

    • “Behavior” changes “state”

The object model

  • Creating an object

    • Declare a variable of the class type

    • Create an object from that variable

    • Often the two steps are in one line...in C++, not in Java

  • Every object gets its own copy of variables and methods

    • Except static objects

  • Type of an object

    • The class of which this object is (inheritance makes this somewhat complicated)

Object-oriented Methodology:
Abstraction

  • Objects expose only those details that are necessary for other objects to know

    • Make themselves appear simpler than they really are

  • Why?

    • To adhere tightly to the expectations from an object

      • If A assumes B can perform a task, B exposes just enough functionality for A to use, no more

    • To hide its own complexity

  • Interface of an object

    • Its “public” face

    • “Interface abstracts implementation”

Object-oriented Methodology:
Abstraction

Object-oriented Methodology:
Abstraction

Object-oriented Methodology:
Abstraction in Design

  • Design the interface for each object first

  • Design one object in detail using only interfaces of the others

    • Details of only one object, interfaces of others

  • Separate the “what” from the “how”

    • “What”: what can an object do? (the interface)

    • “How”: how does the object do it? (details)

    • “What” is used in design: Just care about what other objects do, not how they do it

Object-oriented Methodology:
Standardizing an Abstraction

  • Standardize method names, parameters and output

  • Java: abstract class and interfaces

    • Abstract class: One or more methods are abstract

      • abstract class X

      • abstract public void methodname(…);

    • Interfaces: merely the method signatures

      • interface X

Object-oriented Methodology
Abstraction and Hierarchy

  • Creating a hierarchy...

    • System

      • Sub-systems

        • Components

          • Objects

  • Each level of the hierarchy exposes more details than the one above

  • Provides for the proper level of understanding

Object-oriented Analysis and Design by Grady Booch

Object-oriented Methodology
Encapsulation

  • Combining everything necessary for a cohesive entity (data+operations) into a class

  • Hiding all the (bitter) details within a capsule

  • Classes that represent self-complete entities = good encapsulation

Object-oriented Methodology
Information Hiding

  • Hiding all the details that are not necessary to use the object

    • Other objects

  • Good information hiding: expose only what others can use

    • Information hiding creates the abstraction

  • Information hiding in Java

    • Access modifiers: private, protected, public, default

    • Rule of thumb: make all variables private, only some methods public

    • Accessors and modifiers

  • Interface

    • The set of methods and variables of an object that another object can see and use

Object-oriented Methodology
Cohesion

  • Objects that represent an entity should contain all the functionality within it for that entity (and nothing else)

    • Good encapsulation

  • Bad/Low cohesion: classes that don't completely represent one entity, classes with multiple purposes

    • Classes with no self-sufficient role

      • E.g. classes with methods that change external data only

    • Classes with multiple purposes/roles

      • E.g. class that mixes operations and I/O

Object-oriented Methodology
Decoupling

  • Objects should use each other only when needed

    • Bad/Low decoupling: State of one object being manipulated directly within several objects

  • Effects of Low cohesion and decoupling

    • Compromises functional separation of various objects

    • Code and structure cluttered, complicates testing, maintenance

Characteristics of a Well-Designed Object-Oriented System

  • Objects/components with clear, singular purpose

  • Objects/components with well-defined interactions

  • Good encapsulation

  • Effective use of abstraction through information hiding

  • High cohesion, high decoupling

How does all of this improve our lives?

Advantages of the Object Model

  • Division of the system into well-defined components

  • Well-defined interaction within the system

  • Easier code reuse

  • Easier system integration

  • Applicable to all types of sub-systems

Advantages of the Object Model:
Division into components

  • Semantic division into components (objects)

    • Division into procedures is more awkward

    • Good if cohesion/decoupling is high

  • Independent development and testing

    • Each object is self-sufficient and purposeful

    • Each component can be independently stabilized and made robust

    • If object boundaries are well-defined

  • Resilient to later changes

    • Future upgrades change only some classes

    • Changes are modular and isolated

    • If cohesion/decoupling is high

Advantages of the Object Model:
Well-defined interactions

  • Objects interact with each other by message passing

    • Calling each other methods

    • Method names, signatures decided during design

      • Enforcing this in "structured" design is difficult

  • Good design disciplines interactions

    • Well-defined message passing

    • Effect of change is deterministic and finite

      • If decoupling is high

  • Interaction and cohesion: a balance

Advantages of the Object Model:
Code re-use

  • Using readymade code is always preferable

    • Less time and effort than re-inventing the wheel

    • Less testing, higher reliability

      • Provided code is from a reliable source

  • The basis for all readymade packages and libraries

  • Object re-use more convenient than procedural re-use

    • Objects well-defined components

    • Pick-and-choose re-use possible because of overriding

    • Well-defined interface, more amenable for reuse

Advantages of the Object Model:
Easier System Integration

  • Early dependency detection

    • During analysis and design

  • Concurrency and version control easier

    • Versions of objects instead of files

  • Integration testing is separable and delayed

    • Using “placeholders” is easier to delay integration testing

    • Known message passing interfaces from design

  • Integration process is smoother

    • Well-fitting, “pre-conceptualized” parts

Advantages of the Object Model:
Applicable to all types of subsystems

  • User interface design

  • Data access design

  • Processing workhorses