Identifying and Improving Reusability Based on Coupling Patterns

1
Identifying and Improving Reusability Based on
Coupling Patterns

• Andrea Capiluppi and Cornelia Boldyreff
• Department of Computing and Informatics
• CROSS
• University of Lincoln

2
(No Transcript)
3
(No Transcript)
4
Motivation and Background

• Successful in-house commercial software reuse
• The promise of reusable COTS has been realised,
  including OSS system components LAMP
• Little reuse from OSS projects in OSS projects

5
Reasons for Missing Reuse link in OSS

• Reuse not considered a critical success factor
• Potential for reuse not investigated
• Duplication of effort seen as leading to survival
  of the fittest!
• Independence of OSS projects not invented here
  syndrome

6
Our Objective Reuse Technique for OSS Community

• Assist OSS projects in identifying reusable
  components, so their benefits can be potentially
  realised in future OSS projects.
• Focus on source code folders/subdirectories
• Analysing folder-to-folder coupling as extracted
  via the Doxygen engine

7
(No Transcript)
8
Coupling Types

• Dependency Relationship
• Include Relationship
• Function Call Relationship
• Afferent (in-bound) and Efferent (out-bound)
  Couplings

9
Instability Metric

• I Ce / (Ca Ce) see Gorton and Zhu 2005
• The Instability is the ratio of efferent coupling
  to total coupling and ranges from 0 to 1.
• Low Instability equals high reuse potential.
• Since couplings among folders can vary due to
  number of calls, weighted factors have been
  introduced.

10
Case Studies for Analysis

• Arla (8 years), Mplayer (4.5), XMMS (5), Gaim
  (6.5) all productive with frequent releases and
  thriving development communities
• libc (C Library) and intl (local language) reused
  across these projects

11
Source Folders as Reusable Units

• initial results show that OSS developers actively
  use folders as containers for similarly scoped
  elements and prefer intra-folder linking (on
  average Arla's folders appear less reusable, i.e.
  more unstable).
• However, as folders increase in size over time,
  Ca decreases (architectural erosion).

12
Identifying Reusable Folders

• Table 1 Reusable folders detected via coupling
  analysis shows a total of 14 potentially reusable
  folders
• 6 false positives, but 8 true positives!
• Implemented algorithm does actually select
  folders that are being actively reused (rx reused
  in Arla and tremor in MPlayer)

13
Future Developments

• Refine the selection algorithm consider other
  types of coupling and other reuse characteristics
  to determine reuse potential.
• Our vision a plug-in to widely used OSS
  configuration management systems, e.g. cvs and
  subversion, to automatically run our algorithm
  over the source code folders on check-in and thus
  alert developers to reuse possibilities!