use LoadErrors to detect RSpec version

[tooky]

I had a problem when running cucumber that the Spec constant appeared to be defined even though I was using RSpec 2, and therefore getting a LoadError when trying to load spec/runner/formatter/progress_bar_formatter.

This pull request uses LoadErrors to determine which rspec is loaded. It defaults to 2, and falls back to 1 if there is a LoadError.

Cheers, /tooky

[grosser]

I think the idea is good, but solution could be made cleaner. Rspec2 also defined Spec, to catch errors I suppose, so it seems to me that the check only needs to be reversed to if not defined? RSpec

[grosser]

hmm just gave it a try and defined? Spec is nil, but calling Spec returns RSpec, strange... Can you try Spec::VERSION::STRING and see if it returns something ?

[grosser]

please give the not defined? a try, if it works then this would be a smaller change/fix

[tooky]

hmmm, its weird.

The original breaks running my cucumber features, but I can't seem to find Spec defined anywhere.

The pattern I used here is the same as used in cucumber-rails.

[tooky]

Won't RSpec always be defined though as we are inside an RSpec module at this point?

Doesn't rescuing the LoadError ensure that we are using rspec2 if available, and then trying rspec1 if we don't, and we only throw a LoadError if that isn't available either?

[grosser]

yep good point, we can have a version = defined?(Rspec) ? 2 : 1 line above the module definition, but another straigt-forward test would be nice (e.g. Spec::VERSION::STRING / RSpec::Version::STRING)

Im not a fan of wrapping a bunch of code in a big rescue block, but having the rescue at the top would also work require 'xxx'; version = 1; rescue LoadError; require 'yyy'; version = 2

[tooky]

Yep agreed having a bunch of code in a rescue isn't ideal.

Just wondering if we can have the class definitions in separate files - just require the version 1 or 2 file as needed?

[grosser]

looks pretty clean, ill merge that :)

[tooky]

Great, thanks.