Design Metrics CS 406 Software Engineering I Fall 2001

1
Design MetricsCS 406 Software Engineering IFall
2001

• Aditya P. Mathur

Last update October 23, 2001
2
Metrics

• Metrics are useful in measuring the complexity
  and goodness of a design.

• A large number metrics have been proposed for OO
  designs.

• Some of these have been validated experimentally,
  others are mere proposals or have received little
  or no validation.

3
Effort

• Assumption The effort in developing a class is
  determined by the number of methods.

• Hence the overall complexity of a class can be
  measured as a function of the complexity of its
  methods.

Proposal Weighted Methods per class (WMC)
4
WMC

• Let class C have methods M1, M2, .....Mn.

• How to measure WMC?

5
WMC validation

• Most classes tend to have a small number of
  methods, are simple, and provide some specific
  abstraction and operations.

• WMC metric has a reasonable correlation with
  fault-proneness of a class.

6
Depth of inheritance tree 1

• Depth of a class in a class hierarchy determines
  potential for re-use. Deeper classes have higher
  potential for re-use.

• Inheritance increases coupling. Changing classes
  becomes harder.

• Depth of Inheritance (DIT) of class C is the
  length of the shortest path from the root of the
  inheritance tree to C.

7
Depth of inheritance tree 2

• In case of multiple inheritance DIT is the
  maximum length of the path from the root to C.

• Depth of Inheritance (DIT) of class C is the
  length of the shortest path from the root of the
  inheritance tree to C.

8
DIT evaluation

• Basili et al. study,1995.
• Chidamber and Kemerer study, 1994.
• Most classes tend to be close to the root.
• Maximum DIT value found to be 10.
• Most classes have DIT0.
• DIT is significant in predicting error proneness
  of a class. Higher DIT leads to higher
  error-proneness.

9
DIT evaluation

• Basili et al. study,1995.
• Chidamber and Kemerer study, 1994.

• Most classes tend to be close to the root.

• Maximum DIT value found to be 10.

• Most classes have DIT0.

• DIT is significant in predicting error proneness
  of a class. Higher DIT leads to higher
  error-proneness.

10
Number of children (NOC)

• NOC is the number of immediate subclasses of C.

• Higher values of NOC suggest reuse of the
  definitions in the super-class in a larger number
  of subclasses.

• Higher NOC suggests the extent of influence of a
  class on other elements of a design. Higher
  influence demands higher quality of that class.

11
Validation of NOC

• Classes generally have a small NOC value.
• Vast majority have NOC0.
• Larger NOC value is associated with lower
  probability of detecting faults in that class.

12
Coupling between classes (CBC)

• Class C1 is coupled to class if at least one
  method of C1 uses a method or an instance
  variable of C2.
• Coupling is usually easy to identify though often
  pointers may make it difficult.
• CBC of Ctotal number of other classes to which C
  is coupled.

13
Validation of CBC

• Most classes are self contained and have CBC0.
• Interface classes tend to have higher CBC values.
• CBC is significant in predicting fault-proneness
  of classes.

14
Response for a class (RFC)

• Response set of class C is the total number of
  methods that can be invoked when a message is
  sent to an object of C. This includes all methods
  of C and any methods executed outside of C as a
  result of this message.
• RFC of class C is the cardinality of the
  response set of C.
• Note that even when CBC1 RFC may be high. This
  indicates that the volume of interaction is
  high.

15
Validation of RFC

• Most classes tend to invoke a small number of
  methods (low RFC values).
• Classes for interface objects tend to have larger
  RFC values.
• RFC is very significant in predicting the
  fault-proneness of a class.

16
Lack of cohesion in methods (LCOM)

• Let I1 and I2 denote sets of instance variables
  of methods M1 and M2, respectively, in class C.
• M1 and M2 are considered similar, or cohesive, if
  I1 and I2 are not disjoint.
• Let Q be the set of all cohesive method pairs.
• Let P be the set of all non-cohesive method
  pairs.
• LCOMP - Q if P gt Q, 0 otherwise.

17
LCOM

• A larger number of cohesive pairs implies smaller
  LCOM.
• A high value of LCOM suggests that a class is
  trying to support multiple abstractions. Perhaps
  the class needs to be partitioned into smaller
  and more cohesive classes.
• LCOM is found to be not very significant in
  predicting fault-proneness.

18
Summary

• What are OO metrics?
• Metrics for complexity, coupling, and cohesion