On Feb 6, Robert C. Martin (“Uncle Bob”) gave a talk on “Automated Acceptance Testing: Expressing Requirements as Tests” @ ObjektForum Karlsruhe.
A few statements I found worthwhile for future reference:
- Deployment is a business decision
- Agile is not about getting more done, it’s about getting bad news early
- Don’t let the tool dominate you
- Not a lot of fixture code to write (to enable automated acceptance testing)
- Test one thing only once
Here’s a thought which struck me while Uncle Bob waxed eloquently about Fitnesse and showed the DSL he created for testing Fitnesse:
A textual representation of a program’s or component’s output is the major prerequisite for automated acceptance testability. If you struggle with automated acceptance tests – as we do (CAD system with rich user interactions and complex objects being generated, dealing with complex object lifetime issues – “this can’t be tested automatically”) – focus on generating a simple textual representation of the results & interactions etc. This is the foundation of automated acceptance testability. It will be way easier to think about automated acceptance testing once you have this foundation in place.
Thanks to Robert for giving an energized talk despite being attacked by a flu virus and to andrena objects ag for organizing the ObjektForum event series.