Software refactoring guided by multiple softgoals

1
Software refactoring guided by multiple soft-goals

• Yijun Yu, John Mylopoulos, Eric Yu, Julio C.
  Leite, Linda Lin Liu, University of Toronto
• Erik H. DHollander, University of Ghent

2
1. Refactoring modelling performance vs.
complexity

• real8 A(512,512), B(512,512), C(512,512)
• M L N 512
• do i 1,M
• do j 1, L
• do k 1, N
• C(i,k) C(i,k) A(i,j) B(j,k)
• GOAL apply transformations to speedup the above
  program 20 times without increasing the code
  complexity 4 times

3
Experiment results
Rationale in action gt
4
soft-goalType Topic
5
TYPE TOPIC
operationalization
OPERATION-ALIZED SOFTGOAL
TYPE SUBTOPIC
Topic taxonomy
Decomposition method
ltlabelgt
TYPE TOPIC
Contributioninterdependency
Type taxonomy
SUBTYPE TOPIC
6
Label propagations verify the choices
7
2. Software evolution

• Without efforts, a software is generally
  subjected to
• 1. Continuing changes
• 2. Increasing complexity
• 7. Declining quality in terms of performance
• Lehmann96 Laws of software evolution revisited

8
2. Header project functionality and quality
refactoring shift focus along evolution
9
3. The goal-driven refactoring process

• Modelling non-functional qualities as soft-goals
  Mylopoulos92
• Are they met? Decomposing soft-goals into
  interdependency graph Chung94
• Catalogue the refactoring steps as
  operationalized soft-goals and make selections
  c.f. Tahvildari01, Zou02
• Apply the refactoring steps and propagate the
  labels up to re-assess the soft-goals
  Giorgini02
• Release the product that satisfice the NFRs
• Redo refactoring after functionality changes
  Fowler99

10
4. Summary
SIG
Qualitytype topic taxonomy
1. Modeling
--
-


6. Evolution
Rationale prioritization (bottleneck), metrics,
constraints
2. Satisfiable?
5. Releasing
3. Selection
Operationalizations refactoringsteps
labelpropagation


4. Refactoring