A GraphBased Metamodel for ObjectOriented Software Metrics

1
A Graph-Based Metamodelfor Object-OrientedSoftwa
re Metrics

• Tom Mens (tom.mens_at_vub.ac.be)
• Postdoctoral Fellow Fund for Scientific
  Research (Flanders)
• Vrije Universiteit Brussel, Belgium

2
Goal

• use graphs as an underlying formalism/representati
  on to
• define a formal framework for OO metrics
  independent of
• the programming language (e.g., Smalltalk, Java,
  or C)
• the life-cycle phase (e.g., design or
  implementation)
• identify a minimal set of primitive functions to
  cover an as wide variety of OO metrics as
  possible
• implement open and customisable metrics tools
• easy to add new OO metrics
• easy to incorporate new language features
• easy to use and customise at different levels of
  abstraction

3
Approach

• top-down
• improve existing metrics tools
• CodeCrawler
• SoulMetrics
• bottom-up
• develop generic graph-based formalism
• validate
• on a variety of OO metrics suites
• on a significant number of different programs

4
Tool support CodeCrawler

• a language independent reverse engineering tool
• combines metrics and software visualization
• based on the FAMIX language-independent metamodel
• implemented in VisualWorks 3.0 Smalltalk
• runs on every major platform

5
Tool support CodeCrawler
6
Tool support CodeCrawler
7
Tool support SoulMetrics

• a generic logic-programming-based metrics tool
• defined in SOUL, a logic meta-programming
  environment
• on top of Smalltalk VisualWorks 7
• a collection of logic predicates
• fully integrated in the Smalltalk GUI (browser)
• runs on every major platform

8
Tool support SoulMetrics
9
Tool support SoulMetrics
10
Tool support SoulMetrics
11
Represent software as graphs

• Directed attributed multi-graphs

12
Primitive graph functions

• pred Node ? PF(Node) predecessor nodes
• succ Node ? PF(Node) successor nodes
• fanIn Node ? PF(Edge) incoming edges
• fanOut Node ? PF(Edge) outgoing edges
• path Node ? Node ? PF(Edge) edge paths
• iterative versions predi, pred, pred, succi,
  succ, succ

13
Primitive graph functions

• pred(c4)

14
Primitive graph functions

• succ(c4)

15
Primitive graph functions

• fanIn(c3)

16
Primitive graph functions

• fanOut(c3)

17
Primitive graph functions

• path(c2,c)

18
Qualified graph functions

• pred Node?NodeConstraint?EdgeConstraint ?
  PF(Node)
• idem for predi, pred, pred, succi, succ, succ
• fanIn Node?EdgeConstraint ? PF(Edge)
• idem for fanOut
• path Node?Node?NodeConstraint?EdgeConstraint ?
  PF(Edge)
• begin is defined in terms of pred
• calculates all nodes that have no predecessors
• end is defined in terms of succ
• calculates all nodes that have no successors

19
Qualified graph functions

• fanIn(c,isNoUsesEdge)

20
Qualified graph functions

• pred(c,isClassNode,isUsesEdge)

21
Object-oriented inheritance metrics

• Let n?Node such that isClassNode(n)
• subclasses(n) pred(n,isClassNode,isInheritsEdge
  )
• descendants(n) pred(n,isClassNode,isInheritsEd
  ge)
• leafClasses(n) start(n,isClassNode,isInheritsEd
  ge)
• inheritanceToRoot(n) ? path(n,m,isClassNode,is
  InheritsEdge)
• m?end(n,isClassNode,isInheritsEdge
  )
• NOS(n) subclasses(n)
• NOD(n) descendants(n)
• NOL(n) leafClasses(n)
• DIT(n) average(inheritanceToRoot(n),map(length))

22
Object-oriented inheritance metrics

• subclasses(c) NOS(c)2

23
Object-oriented inheritance metrics

• descendants(c) NOD(c)4

24
Object-oriented inheritance metrics

• leafClasses(c) NOL(c)2

25
Object-oriented inheritance metrics

• inheritanceToRoot(c2) DIT(c2)2

26
Ratio Metrics

• MethodHidingFactor(n)
• 1 Ratio(succ,isMethodNode,isPublicNode,isConta
  insEdge)(n)
• 1- succ(n,isMethodNode?isPublicNode,isContainsE
  dge)
• succ(n,isMethodNode,isContainsEd
  ge)
• AbstractSubclassRatio(n)
• 1 Ratio(pred,isClassNode,isAbstractNode,isInhe
  ritsEdge)(n)
• LeafclassRatio(n)
• 1 Ratio(pred,isClassNode,isInheritanceLeaf,is
  InheritsEdge)(n)