On Sun, 17 Feb 2008 12:40:31 -0500, David Chelimsky wrote: > On Feb 17, 2008 11:17 AM, Steve > <[EMAIL PROTECTED]> wrote: >> >> In the example I actually have "include ApplicationHelper" in there, >> and am using a method "row_class_for_idx" in the spec. It works. I'm >> just wondering if that's bad form or not. > > You may want to give this a read: > http://blog.jayfields.com/2008/02/testing-expect-literals.html. > > Please do write back here if you do and let us know what you think.
By and large I agree with what Jay is saying, and I try to use literals as much as possible already. I don't try to make my specs DRY by any means, and do in fact have much that is redundant because it improves readability. I do have some arguments for constants and variable values though. For constants if you have the constant spec'd to ensure that it itself is the correct value, then why not use the constant in your specs? You're not saying that the code being tested has to use the constant, just that the algorithm comes out correct. If the spec for the constant fails, you know that it's not what it's supposed to be. If you don't layer your specs like this, then I could see how it would be a problem, but I personally spec my constants. For variable values, such as using a method. Sometimes I just don't care. The method being called is trivial. The example I presented above is good. Right now row_class_for_idx returns "" for even, and "alt" for odds. Let's say in the future I wanted it to be "even", and "odd" for the class names respectively. It's a very simple requirement, used commonly to make sure the lists are going to display properly. It seems like something that's just tedious to go back and change. Also since row_class_for_idx has it's own specs, I know that it will be putting out what I want. _______________________________________________ rspec-users mailing list rspec-users@rubyforge.org http://rubyforge.org/mailman/listinfo/rspec-users
