I've been working on a Rails project with one other developer; he was using Test::Unit, and I was using RSpec. That works OK for a while, but obviously it starts causing pain when you have to check in two places to see if a piece of code is properly tested/spec'd, you can't use TextMate shortcuts to switch back and forth between code and test, you have to duplicate shared behaviors/test helpers, etc.
So when we brought in a consulting team to add some manpower, we realized we had to switch to a single framework. This is a team that's fairly experienced with Rails and active in the Rails community, but was quite opposed to choosing RSpec. Here are the arguments I heard against unifying on RSpec: * Test::Unit is ubiquitous. Everyone knows it. This is hard to counter; it comes with Rails and is the default. Same reason many people use Prototype even though JQuery/dojo might suit them better. * For that reason, it's a lot easier to find examples of "how to do something" in Test::Unit than in RSpec. That's true; several times I've had a bit of code that didn't fall nicely into the MVC hierarchy, and I wasn't sure how to build up the right context to test it in. If I were using Test::Unit, I could just copy the equivalent tests from Rails core, but using RSpec I had to roll my own. * RSpec is BDD (hand-waving new different troublesome); we do TDD. We've covered that ground on this list many times; BDD is an extension and interpretation of TDD, not some newfangled crackpot theory. But people don't know that. * The team had in fact investigated RSpec a few months ago, and decided they didn't like it. Some of what they didn't like has been fixed in 1.0, but of course people aren't going to come running to re-examine each release, so the bitter taste remained: * #context was defined on Kernel. Not sure if that's still true for #describe. * Not compatible with tools that expect Test::Unit output. This would(could) be fixed with the runner integration that's been discussed. * Wasn't compatible with mocha/FlexMock. Fixed now. But again, it came back to ubiquity, which is a pretty hard problem to overcome. Seems to me that the best way to get RSpec adopted is to find some more visible, prolific plugin programmers and evangelize them to start using RSpec, so it's not some "neat fringe thing", but a solid, respectable alternative to Test::Unit. Jay Levitt _______________________________________________ rspec-users mailing list rspec-users@rubyforge.org http://rubyforge.org/mailman/listinfo/rspec-users
