TDD: The Good, The Bad, and the Ugly

Jeffery Palermo has a very good posting about the good, the bad, and the ugly of a TDD project. Our project has experience many of the same effects. We'll worth a read.

http://codebetter.com/blogs/jeffrey.palermo/archive/2006/04/12/142757.aspx

I would love to hear more about how they integrate FitNesse into their project. I feel our team is on the bleeding edge by using FitNesse with .NET but now that we are using FitNess we can't live without it.

Questions:
Who writes the Fit test? Our customers don't speak “FIT Table“ yet, but I'm working on it.
How do you integrate the FitNesse tests into the CruiseControl.NET build process?
Is there a easy way to debug FIT tests?

posted @ Monday, April 17, 2006 9:31 PM

Print

Comments on this entry:

# re: TDD: The Good, The Bad, and the Ugly

Left by Jean-Paul S. Boodhoo at 4/27/2006 8:23 PM
Gravatar

On the last project I rolled off from, we developed a framework that allowed the business users to write the fit tests in Excel spreadsheets. We wrote a component that would run and execute fit using the excel spreadsheets as input. This worked great as it allowed the customers to write the fit tests for us. When they needed more descriptive capabilities in their "tests" we would extend the framework as necessary. Of course writing the tests is one thing, but for each new "fixture" in a fit test, a new class has to be written to handle the fixture. You end up with a bunch of reusable fixtures that the clients become accustomed to and compose together to create more complicated tests. At the end of the day, it was just delegating to fit, so it was pretty simple. On our "big bertha build" fit ran as part of the cruise control build. Of course, before a checkin, all developers had to run the fit tests on their local machine to ensure that they had not messed up any functionality. As far as debugging, if you actually had to spend time in the debugger, you could attach yourself to the fit process, set a breakpoint and let it run. More often than not, we just logged like crazy, as it was a much more efficient debug tool. On our big build, we also ran WATIR tests to exercise the UI. This ensured that we had full coverage from UI all the way to mapping.

Comments have been closed on this topic.