Measuring the Effect of Test-Driven Development

Keith Braithwaite (Zuhlke)

Discovery Sessions · Developing, Testing

Tuesday, 10:30, 1 hour 30 minutes | Meeting Room 5

Download Presentation Download Handout

Test-driven Devlopment is perhaps the most widely adpoted technical practice in the Agile development toolkit. TDD can be a difficult practice to introduce, requiring developer training and/or mentoring, changes to team functions and behaviour, project planning and so forth. It would be easier to justify these costs if the benefits of TDD could be clearly demonstrated. Many practitioners report that TDD (together with refactoring) has a profound qualitative impact on the design of their code, and not just on defect rates and other quality-control measures. The presenter has found some evidence that some of these qualitative influences on design of TDD can be quantified by examining certain statistical properties of a body of (Java) code. If the hypothesis that the qualitative difference between TDD and non-TDD code can be objectively measured is confirmed then it will be possible to track the effect of introducing TDD, meaure the effectivness of a TDD programme and place TDD exercises on a firm footing. This would improve developers' confidence in their practice, provide corrective feedback to that practice and provide evidence to use in influencing figures like project managers. But to confirm it requires that some empirical study be done on TDD and non-TDD codebases. This and other conference sessions form part of that research--and an opportunity for conference attendees to take part in some genuine science in the field of computers.