Verification

1
Verification Validation of OOMeter

• Presented By Malik M. Umar

2
Objectives

• Make a literature survey on the metrics
  implemented in OOMeter.
• Install Junit testing framework and be familiar
  with it
• Obtain OOMeter source code
• Get some test java source code and XMI files.
• Manually compute or otherwise obtain the metric
  values on the item in (3) above

3
Objectives

• Write Junit test cases with assertions comparing
  your manually obtained values and values obtained
  from other tools with values computed by OOMeter
• Report your findings
• If your findings detect errors in OOMeter's
  computation, suggest a fix.

4
Metrics Literature Survey

• Number of Children
• Depth of Inheritance Tree
• Cohesion among Methods in a Class
• Loose Class Cohesion
• Tight Class Coupling
• Lack of Cohesion in Methods 1

5
Metrics Literature Survey

• Lack of Cohesion in Methods 2
• Lack of Cohesion in Methods 3
• Lack of Cohesion in Methods 4
• Lack of Cohesion in Methods 5
• Coupling Between Object Classes

6
Observations

• Data Structure Coupling to Metric Code
• Lack of Modularity in Metric Implementations
• Database Platform Dependence
• Merging of Core with Frill Functions

7
Recommendation for Data Structure Coupling to
Metric Code

• Solution
• Abstract Data Layer

8
Recommendation forLack of Modularity in Metric
Implementations
9
Recommendation forLack of Modularity in Metric
Implementations

• Solution Define a Metric Writing Standard and
  Design Pattern
• The metric interface should include
  preCalculate() and postCalculate() methods for
  initialization and cleanup operations.
• The calculate method should only be used for
  starting the metric calculation and the business
  logic of metric calculation should be place in
  one or more public methods.
• Code to store the metric calculation results
  should not be in the same class as the metric
  implementation class, rather should be moved to a
  separate helper class.
• Any formulas/logic used to calculate the metrics
  should be included in the metric as comments.

10
Recommendation forDatabase Platform Dependence

• No More SQL
• No more SQL queries to implement metrics. All
  metrics should be developed using java code as
  apposed to SQL.
• Utilize a native XML database developed in Java
  such as eXists (http//exist.sourceforge.net/) or
  Java based RDBMS e.g. cloudscape.
• Utilize XML based files to store parsing results
  and metric results.
• Use EJBs for the data tier and allow the user to
  select the relevant database at deployment-time

11
Recommendation forMerging of Core with Frill
Functions

• Two Projects Are Better Then One
• Create the core functionality of the application,
  accessible through a command-line interface.
  (Examples of such projects are Ant and, MySQL)
• Develop a GUI based application that uses the
  core application and provides a user friendly
  interface.

12
Recommendation forMerging of Core with Frill
Functions
13
Questions?