3 ms·

A bad architecture and bad code structure can be achieved without TDD. I don't think well thought out design precludes unit testing. I don't think producing ba

A bad architecture and bad code structure can be achieved without TDD. I don't think well thought out design precludes unit testing. I don't think producing bad architecture/bad code structure is goal of TDD and I suspect it's something to be thinking about regardless of whether you use TDD. If you're having to contort your code or change it just to make a test pass, that's probably a smell. It just so happens that the SOLID principles (distinct from TDD) are beneficial for TDD and you would probably consider those to create a 'good architecture' and 'code structure'. http://en.wikipedia.org/wiki/SOLID_%28object-oriented_design%29 http://en.wikipedia.org/wiki/SOLID_%28object-oriented_design...

dozzie 12y ago
> A bad architecture and bad code structure can be achieved without TDD. Of course. But TDD seems to be applied instead of good architecture. The latter is way, way more important than some tests. > I don't think well thought out design precludes unit testing. It doesn't, but the latter is often applied without the former, resulting in claims "it has many tests, so it must be good code".
- axanoeychron 12y ago
  In that example I agree with you. I've seen my fair share of bad design with good test coverage. Then I have also seen my fair share of good architectures with no test coverage. When I see a good architecture with effective testing - that's just awesomeness. I think this is something to endeavour for. I'm not sure how you solve the problem of developers creating bad architectures with heavy test coverage but I don't think pure abandonment of TDD is necessarily the answer (for the alternative which is bad code with no coverage). I do not subscribe to the reasoning that Tests immediately cause a bad design - they definitely influence it but they are not the primary cause.
  - dozzie 12y ago
    > I do not subscribe to the reasoning that Tests immediately cause a bad design Me neither. But tests are given much, much more attention than they are worth. Tests mainly make up for lack of clean structure and good tooling for static analysis. Strong static typing is worth attention. Static analysers and code proof assistants are worth attention. Simple design that make whole class of errors impossible is worth it. Tests are not worth that much in this neighbourhood. Having plenty of tests is OK when the code base is fragile and it's not clear how to make it robust. Tests are a tool of last resort, if all thinking fails to produce something bug-resistant.

dozzie 12y ago

> A bad architecture and bad code structure can be achieved without TDD. Of course. But TDD seems to be applied instead of good architecture. The latter is way, way more important than some tests. > I don't think well thought out design precludes unit testing. It doesn't, but the latter is often applied without the former, resulting in claims "it has many tests, so it must be good code".

axanoeychron 12y ago
In that example I agree with you. I've seen my fair share of bad design with good test coverage. Then I have also seen my fair share of good architectures with no test coverage. When I see a good architecture with effective testing - that's just awesomeness. I think this is something to endeavour for. I'm not sure how you solve the problem of developers creating bad architectures with heavy test coverage but I don't think pure abandonment of TDD is necessarily the answer (for the alternative which is bad code with no coverage). I do not subscribe to the reasoning that Tests immediately cause a bad design - they definitely influence it but they are not the primary cause.
- dozzie 12y ago
  > I do not subscribe to the reasoning that Tests immediately cause a bad design Me neither. But tests are given much, much more attention than they are worth. Tests mainly make up for lack of clean structure and good tooling for static analysis. Strong static typing is worth attention. Static analysers and code proof assistants are worth attention. Simple design that make whole class of errors impossible is worth it. Tests are not worth that much in this neighbourhood. Having plenty of tests is OK when the code base is fragile and it's not clear how to make it robust. Tests are a tool of last resort, if all thinking fails to produce something bug-resistant.