Toward Effective Deployment of Design Patterns for Software Extension A Case Study

1
Toward Effective Deployment of Design Patterns
for Software Extension A Case Study

• T.H. Ng1, S.C. Cheung2, W.K. Chan2, and Y.T. Yu1
• 1Department of Computer Science, City University
  of Hong Kong
• 2Department of Computer Science and Engineering,
  Hong Kong University of Science and Technology

ICSE-WoSQ06, May 21 2006
2
Outline

• Introduction
• Case Study
• Related Work
• Empirical Experimentation
• Conclusion

3
Introduction

• Design patterns by Gamma et al. 8
• Proven solutions to recurring problems 5
• Common belief
• Using design patterns leads to extensible
  software. Is that conjecture legitimate?

5 M.P. Cline, The Pros and Cons of Adopting
and Applying Design Patterns in the Real
World, Comm. ACM, Vol. 39, No. 10, October
1996, pp. 47-49.
8 E. Gamma, R. Helm, R. Johnson, and J.
Vlissides, Design Patterns Elements of
Reusable Object-Oriented Software.
Addison-Wesley, 1995.
4
Challenge

• Difficulty of validating the conjecture
• Requires an objective measurement of
  extensibility, which is still an open
  controversial concept

5
Our Approach

• Goal
• Whether this is any relationship between the
  pattern theme and the Open-Closed principle
• Pattern theme
• Common ideas of patterns
• Open-Closed principle 13
• Advocated by many to achieve extensible software

13 B. Meyer, Object-Oriented Software
Construction. Prentice Hall, 1988.
6
Proposed Scheme

• Two phases
• Mark code fragments as reusable or changeable
• Check structural conformance to the Open-Closed
  principle

7
Pattern Theme

• A pattern requires all changeable code fragments
  to be encapsulated in specified design modules,
  such as concrete subclasses in an inheritance
  hierarchy

8
Open-Closed Principle

• The Principle
• Software entities should be open for extension,
  but closed for modification 23
• Conformance to the principle
• Changeable code fragments are in distinct
  partitions from the reusable ones

23 B. Meyer, Object-Oriented Software
Construction. Prentice Hall, 1988.
9
Outline

• Introduction
• Case Study
• Related Work
• Empirical Experimentation
• Conclusion

10
Case Study

• Multi-user Calendar Manager (MCM)
• Functionalities
• Appointment Scheduling
• Display excessive verbose information for error
  reporting. Three verbose modes are supported
  Null, Exception, and Debug
• Anticipation of new verbose modes in the future
• Deployment of the State pattern
• Goal To prepare for supporting new verbose modes
  later

11
Case Study
Figure 1 State pattern deployment in MCM (in UML
notation)
12
An Undesirable Scenario
Specific to Debug verbose mode
Specific to Debug verbose mode
Specific to Debug verbose mode
Specific to Exception verbose mode
Figure 2 A client procedure that uses both the
State pattern and some unclean statements
(S1-S4)
13
A Undesirable Scenario

• Reusable and changeable statements for the
  support of new verbose modes
• Each statement specific to an existing verbose
  mode, such as S1, S2, S3, and S4, is changeable.
• Each statement not specific to any existing
  verbose mode is reusable
• Violation of Open-Closed Principle
• The implementation of the operation in Figure 2
  comprises both reusable and changeable statements
  for future support of new verbose modes

14
Outline

• Introduction
• Case Study
• Related Work
• Empirical Experimentation
• Conclusion

15
Related Work

• Structural and behavioral conformance 720
• Our case study
• Structural and behavioral conformance is
  satisfied
• The Open-Closed principle is violated

7 R.B. France, D.K. Kim, S. Ghosh, and E. Song,
A UML-Based Pattern Specification Technique,
IEEE Trans. Software Eng., Vol 30, No. 3, 2004,
pp. 193-206..
20 N. Soundarajan, and J.O. Hallstrom,
Responsibilities and Rewards Specifying Design
Patterns, ICSE 2004, Edinburgh, pp. 666-675.
16
Outline

• Introduction
• Case Study
• Related Work
• Empirical Experimentation
• Conclusion

17
Empirical Study

• Null Hypotheses
• When the pattern theme is not satisfied, there is
  at least a 20 chance where open-closed
  conformance is followed in a pattern-based
  program
• When the pattern theme is satisfied, there is at
  least a 20 chance where open-closed conformance
  is violated in a pattern-based program

18
Empirical Study

• Controlled Experiment
• 98 subjects individually developing the MCM
  program
• Materials
• Program requirements
• Functional correctness
• The deployment of the State pattern
• The satisfaction of the pattern theme
• Issues
• Subjects are unaware of the Open-Closed principle

19
Empirical Study Results
40 (40.82) Population satisfying the theme of
patterns
43 (43.88) Population establishing conformance
to the Open-Closed Principle
98 (100) Total population, all deployed the
State patterns as the design in Figure 1
20
Rejecting Hypothesis (a)

• Violating the pattern theme likely leads to the
  violation of conformance to the Open-Closed
  principle

40 (40.82) Population satisfying the theme of
patterns
43 (43.88) Population establishing conformance
to the Open-Closed Principle
Only 3 cases where the Open-Closed principle is
followed when the pattern theme is violated
98 (100) Total population, all deployed the
State patterns as the design in Figure 1
21
?2 Test on the Hypothesis

• a denotes whether the Open-Closed principle is
  achieved
• i denotes the random variable the condition of
  a
• Oi denotes the observed frequency Ei denotes
  the expected frequency, as implied by the
  hypothesis with a 20 chance of achieving
  open-closed conformance when the proposed
  criterion for deployed patterns is not satisfied.
  
• Please be informed that there are 40
  participants satisfying the pattern theme so the
  total observed frequencies are therefore 58
  (98-40).

22
Rejecting Hypothesis (b)

• Satisfying the pattern theme leads to conformance
  to the Open-Closed principle

40 (40.82) Population satisfying the theme of
patterns
43 (43.88) Population establishing conformance
to the Open-Closed Principle
98 (100) Total population, all deployed the
State patterns as the design in Figure 1
23
Outline

• Introduction
• Case Study
• Related Work
• Empirical Experimentation
• Conclusion

24
Conclusion

• Summary
• Research Question
• Using design patterns leads to extensible
  software. Is this conjecture legitimate?
• Experimental results
• Yes, if the pattern theme is satisfied.
• Satisfying the pattern theme is a sufficient and
  close to necessary condition to follow the
  Open-Closed principle
• Future work
• Automate checking of pattern theme
• Study other forces and tradeoffs influencing the
  software design of extensible software

25
Questions and Answers