On Sun, Mar 24, 2013 at 11:40 AM, Michal Svoboda <pht@...> wrote:
**
I think thats not such a bad approach. We use it quite a lot:
1. write dirty code of what you "want to do" (aka "spike")
2. observe the result
3. write automatic tests on anything that you verified manually
4. refactor/test-drive "mercilessly" and write new tests as you go
I think that the term for this is "Characteristic Tests", I was
first introduced to it in Michael Feathers Work.
but I'm not sure if it has a special section is his "Working with Legacy
system": book or was it in a later video session i saw.
A very useful technique in general, and should be an excellent idea in this
case.
Lior
_
