Friday, February 29, 2008

How to measure quality of software product

Objectives
Provide status on a new Software Product Quality
Measurement standard and its connection to CMMI
Provide ideas on how to get started with Software Product
Quality Measurement today

Outline
Background and Overview
Concepts and Models
Software Product Quality Measurement
Summary

Achieving Quality Software
Requires planning and intentional design
More than achieving the desired functionality
Must explicitly attend to both functional and non-functional
requirements
Need to verify all requirements are being met throughout
the life cycle

CMMI Definition for Quality Requirements
The phrase “quality and process-performance
objectives” covers objectives and requirements for
product quality, service quality, and process
performance. Process performance objectives include
product quality.

Requirements Development
This process area describes three types of requirements:
• customer requirements (quality in use)
• product requirements (external quality attributes)
• product-component requirements (internal quality
attributes)
Taken together, these requirements address the needs of
relevant stakeholders, including those pertinent to various
product life-cycle phases (e.g., acceptance testing criteria)
and product attributes (e.g., safety, reliability,
maintainability).
Requirements also address constraints caused by the
selection of design solutions (e.g., integration of
commercial off-the-shelf products).

Requirements Development Goals
SG 1 Develop Customer Requirements
Stakeholder needs, expectations, constraints,
and interfaces are collected and translated into
customer requirements.
SG 2 Develop Product Requirements
Customer requirements are refined and
elaborated to develop product and productcomponent
requirements.
SG 3 Analyze and Validate Requirements
The requirements are analyzed and validated,
and a definition of required functionality is
developed.

Process Management and Performance
The organization’s process needs and objectives
cover aspects that include the following:
• characteristics of the processes
• process performance objectives, such as time to
market and product quality
• process effectiveness
A quantitatively managed process is institutionalized
by doing the following:
• controlling the process using statistical and other
quantitative techniques such that product quality,
service quality, and process performance attributes

Key Points in Relationship of CMMI
and ISO 9126/25000

CMMI takes a total life cycle view and is inclusive in its
approach to requirements development.
Requirements development explicitly seeks to have
the developer consider quality requirements.
Project and Process Management processes explicitly
consider product quality as process performance
objectives.
Neither the standard nor CMMI endorses a unidimensional
view of quality.

Product Quality Requirements are transformed into
designs and implemented via the Technical Solution
and Product Integration process areas.
The implementation of Product Quality Requirements
are monitored and confirmed via the Project
Management, Verification, and Validation process
areas.
CMMI acknowledges the need for interaction and
perhaps iteration among the related process areas to
satisfactorily identify, specify, and address Product
Quality Requirements.
are measurable and controlled throughout the
project (internal and external quality measures and
criteria)

Relating Requirements, Evaluation,
and Measurement



Reference: http://www.sei.cmu.edu/sema/presentations/esepg.pdf

Observer and Observable

Here's the problem: You're designing a program that will render data describing a three-dimensional scene in two dimensions. The program must be modular and must permit multiple, simultaneous views of the same scene. Each view must be able to display the scene from a different vantage point, under different lighting conditions. More importantly, if any portion of the underlying scene changes, the views must update themselves.
None of these requirements presents an insurmountable programming challenge. If the code that handles each requirement had to be written de novo, however, it would add significant work to the overall effort. Fortunately, support for these tasks is already provided by the Java class library in the form of interface Observer and class Observable. The functionalities of these two were inspired, in part, by the requirements of the Model/View/Controller architecture.
The Model/View/Controller architecture
The Model/View/Controller architecture was introduced as part of the Smalltalk-80 version of the Smalltalk programming language (a popular object-oriented programming language invented by Alan Kay). The Model/View/Controller architecture was designed to reduce the programming effort required to build systems that made use of multiple, synchronized presentations of the same data. Its central characteristics are that the model, the controllers, and the views are treated as separate entities, and that changes made to the model should be reflected automatically in each of the views.
In addition to the program example described in the opening paragraph above, the Model/View/Controller architecture may be used for projects such as the following:
• A graph package that contains simultaneous bar-chart, line-chart, and pie-chart views of the same data
• A CAD system, in which portions of the design can be viewed at different magnifications, in different windows, and at different scales

Figure 1 illustrates the Model/View/Controller architecture in its most general form. There is one model. Multiple controllers manipulate the model; multiple views display the data in the model, and change as the state of the model changes.



Figure 1: The Model/View/Controller architecture
The Model/View/Controller architecture has several benefits:
• There is a clearly defined separation between components of a program -- problems in each domain can be solved independently.
• There is a well defined API -- anything that uses the API properly can replace either the model, the view, or the controller.
• The binding between the model and the view is dynamic -- it occurs at run time, rather than at compile time.

By incorporating the Model/View/Controller architecture into a design, pieces of a program can be designed separately (and designed to do their job well) and then bound together at run time. If a component is later deemed to be unsuitable, it can be replaced without affecting the other pieces. Contrast that scenario with the monolithic approach typical of many quick-and-dirty Java programs. Often a frame contains all of the state, handles all events, does all of the calculations, and displays the result. Thus, in all but the simplest of such systems, making changes after the fact is not trivial
Defining the parts
The model is the object that represents the data in the program. It manages the data and conducts all transformations on that data. The model has no specific knowledge of either its controllers or its views -- it contains no internal references to either. Rather, the system itself takes on the responsibility of maintaining links between the model and its views and notifying the views when the model changes.
The view is the object that manages the visual display of the data represented by the model. It produces the visual representation of the model object and displays the data to the user. It interacts with the model via a reference to the model object itself.
The controller is the object that provides the means for user interaction with the data represented by the model. It provides the means by which changes are made, either to the information in the model or to the appearance of the view. It interacts with the model via a reference to the model object itself.
At this point a concrete example might be helpful. Consider as an example the system described in the introduction.



Figure 2: Three-dimensional visualization system
The central piece of the system is the model of the three-dimensional scene. The model is a mathematical description of the vertices and the faces that make up the scene. The data describing each vertex or face can be modified (perhaps as the result of user input or a scene distortion or morphing algorithm). However, there is no notion of point of view, method of display (wireframe or solid), perspective, or light source. The model is a pure representation of the elements that make up the scene.
The portion of the program that transforms the data in the model into a graphical display is the view. The view embodies the actual display of the scene. It is the graphical representation of the scene from a particular point of view, under particular lighting conditions.
The controller knows what can be done to the model, and implements the user interface that allows that action to be initiated. In this example, a data entry control panel might allow the user to add, modify, or delete vertices and faces.
Observer and Observable
The Java programming language provides support for the Model/View/Controller architecture with two classes:
• Observer -- any object that wishes to be notified when the state of another object changes
• Observable -- any object whose state may be of interest, and in whom another object may register an interest

These two classes can be used to implement much more than just the Model/View/Controller architecture. They are suitable for any system wherein objects need to be automatically notified of changes that occur in other objects.
Typically, the model is a subtype of Observable and the view is a subtype of observer. These two classes handle the automatic notification function of the Model/View/Controller architecture. They provide the mechanism by which the views can be automatically notified of changes in the model. Object references to the model in both the controller and the view allow access to data in the model
Observer and Observable functions
The following are code listings for observer and observable functions:
Observer

public void update(Observable obs, Object obj)
Called when a change has occurred in the state of the observable.

Observable

public void addObserver(Observer obs)
Adds an observer to the internal list of observers.
public void deleteObserver(Observer obs)
Deletes an observer from the internal list of observers.
public void deleteObservers()
Deletes all observers from the internal list of observers.
public int countObservers()
Returns the number of observers in the internal list of observers.
protected void setChanged()
Sets the internal flag that indicates this observable has changed
state.
protected void clearChanged()
Clears the internal flag that indicates this observable has
changed state.
public boolean hasChanged()
Returns the boolean value true if this observable has changed
state.
public void notifyObservers()
Checks the internal flag to see if the observable has changed
state and notifies all observers.
public void notifyObservers(Object obj)
Checks the internal flag to see if the observable has changed
state and notifies all observers. Passes the object specified in the
parameter list to the notify() method of the observer.

How to use interface Observer and class Observable
The following section describes in detail how to create a new observable class and a new observer class, and how to tie the two together.
Extend an observable
A new class of observable objects is created by extending class Observable. Because class Observable already implements all of the methods necessary to provide the observer/observable behavior, the derived class need only provide some mechanism for adjusting and accessing the internal state of the observable object.
In the class ObservableValue listing below, the internal state of the model is captured by the integer n. This value is accessed (and, more importantly, modified) only through public accessors. If the value is changed, the observable object invokes its own setChanged() method to indicate that the state of the model has changed. It then invokes its own notifyObservers() method in order to update all of the registered observers

import java.util.Observable;
public class ObservableValue extends Observable
{
private int n = 0;
public ObservableValue(int n)
{
this.n = n;
}
public void setValue(int n)
{
this.n = n;
setChanged();
notifyObservers();
}
public int getValue()
{
return n;
}
}



Implement an observer
A new class of objects that observe the changes in state of another object is created by implementing the Observer interface. The Observer interface requires that an update() method be provided in the new class. The update() method is called whenever the observable changes state and announces this fact by calling its notifyObservers() method. The observer should then interrogate the observable object to determine its new state, and, in the case of the Model/View/Controller architecture, adjust its view appropriately

In the following class TextObserver listing, the notify() method first checks to ensure that the observable that has announced an update is the observable that this observer is observing. If it is, it then reads the observable's state, and prints the new value.

import java.util.Observer;
import java.util.Observable;
public class TextObserver implements Observer
{
private ObservableValue ov = null;
public TextObserver(ObservableValue ov)
{
this.ov = ov;
}
public void update(Observable obs, Object obj)
{
if (obs == ov)
{
System.out.println(ov.getValue());
}
}
}



Tie the two together
A program notifies an observable object that an observer wishes to be notified about changes in its state by calling the observable object's addObserver() method. The addObserver() method adds the observer to the internal list of observers that should be notified if the state of the observable changes.
The example below, showing class Main, demonstrates how to use the addObserver() method to add an instance of the TextObserver class (see the TextObserver listing above) to the observable list maintained by the ObservableValue class (see the ObservableValue listing above).


public class Main
{
public Main()
{
ObservableValue ov = new ObservableValue(0);
TextObserver to = new TextObserver(ov);
ov.addObserver(to);
}
public static void main(String [] args)
{
Main m = new Main();
}
}



How it all works together
The following sequence of events describes how the interaction between an observable and an observer typically occurs within a program.
1. First the user manipulates a user interface component representing a controller. The controller makes a change to the model via a public accessor method -- which is setValue() in the example above.
2. The public accessor method modifies the private data, adjusts the internal state of the model, and calls its setChanged() method to indicate that its state has changed. It then calls notifyObservers() to notify the observers that it has changed. The call to notifyObservers() could also be performed elsewhere, such as in an update loop running in another thread.
3. The update() methods on each of the observers are called, indicating that a change in state has occurred. The observers access the model's data via the model's public accessor methods and update their respective views.

A demonstration
The example in Figure 3 demonstrates how observables and observers typically work together in the Model/View/Controller architecture.




Figure 3: A better example
Like the model in the class ObservableValue listing shown previously, the model in this example is very simple. Its internal state consists of a single integer value. The state is manipulated exclusively via accessor methods like those in class ObservableValue. The code for the model is
import java.util.Observable;

public class ObservableValue extends Observable
{
private int nValue = 0;
private int nLow = 0;
private int nHigh = 0;

public ObservableValue(int nValue, int nLow, int nHigh)
{
this.nValue = nValue;
this.nLow = nLow;
this.nHigh = nHigh;
}

public void setValue(int nValue)
{
this.nValue = nValue;

setChanged();
notifyObservers();
}

public int getValue()
{
return nValue;
}

public int getLowerBound()
{
return nLow;
}

public int getHigherBound()
{
return nHigh;
}
}


Initially, a simple text view/controller class was written. The class combines the features of both a view (it textually displays the value of the current state of the model) and a controller (it allows the user to enter a new value for the state of the model). The code is
import java.awt.*;
import java.util.Observer;
import java.util.Observable;

public class TextObserver extends Frame implements Observer
{
private ObservableValue ov = null;

private TextField tf = null;
private Label l = null;

private int nLow = 0;
private int nHigh = 0;

public TextObserver(ObservableValue ov)
{
super("Text Observer Tool");

this.ov = ov;

setLayout(new GridLayout(0, 1));

nLow = ov.getLowerBound();
nHigh = ov.getHigherBound();

tf = new TextField(String.valueOf(ov.getValue()));

add(tf);

l = new Label();

add(l);

pack();
show();
}

public boolean action(Event evt, Object obj)
{
if (evt.target == tf)
{
int n = 0;

boolean boolValid = false;

try
{
n = Integer.parseInt(tf.getText());
boolValid = true;
}
catch (NumberFormatException nfe)
{
boolValid = false;
}

if (n < nLow || n > nHigh)
{
boolValid = false;
}

if (boolValid)
{
ov.setValue(n);

l.setText("");
}
else
{
l.setText("invalid value -- please try again...");
}

return true;
}

return false;
}

public boolean handleEvent(Event evt)
{
if (evt.id == Event.WINDOW_DESTROY)
{
ov.deleteObserver(this);

dispose();

return true;
}

return super.handleEvent(evt);
}

public void update(Observable obs, Object obj)
{
if (obs == ov)
{
tf.setText(String.valueOf(ov.getValue()));
}
}
}


Instances of this view can be created by pressing the upper button in Figure 3.
By designing the system using the Model/View/Controller architecture (rather than embedding the code for the model, the view, and the text controller in one monolithic class), the system is easily redesigned to handle another view and another controller. In this case, a slider view/controller class was written. The position of the slider represents the value of the current state of the model and can be adjusted by the user to set a new value for the state of the model. The code is
import java.awt.*;
import java.util.Observer;
import java.util.Observable;

public class ScrollObserver extends Frame implements Observer
{
private ObservableValue ov = null;

private Scrollbar sb = null;

public ScrollObserver(ObservableValue ov)
{
super("Scroll Observer Tool");

this.ov = ov;

setLayout(new GridLayout(0, 1));

sb = new Scrollbar(Scrollbar.HORIZONTAL,
ov.getValue(), 10,
ov.getLowerBound(),
ov.getHigherBound());

add(sb);

pack();
show();
}

public boolean handleEvent(Event evt)
{
if (evt.id == Event.WINDOW_DESTROY)
{
ov.deleteObserver(this);

dispose();

return true;
}
else if (evt.id == Event.SCROLL_LINE_UP)
{
ov.setValue(sb.getValue());
return true;
}
else if (evt.id == Event.SCROLL_LINE_DOWN)
{
ov.setValue(sb.getValue());
return true;
}
else if (evt.id == Event.SCROLL_PAGE_UP)
{
ov.setValue(sb.getValue());
return true;
}
else if (evt.id == Event.SCROLL_PAGE_DOWN)
{
ov.setValue(sb.getValue());
return true;
}
else if (evt.id == Event.SCROLL_ABSOLUTE)
{
ov.setValue(sb.getValue());
return true;
}

return super.handleEvent(evt);
}

public void update(Observable obs, Object obj)
{
if (obs == ov)
{
sb.setValue(ov.getValue());
}
}
}


Instances of this view object can be created by pressing the bottom button in Figure 3.
Conclusion
Coming up next month is an introductory "how-to" on graphics that should prepare the way for future columns on advanced topics such as animation and custom components. As always, I am interested in hearing from you. Send me mail if you have a suggestion for a future column or a question you would like addressed.