[rspec-users] given-when-then-and syntax support in Rspec

David Chelimsky dchelimsky at gmail.com
Mon Mar 28 00:10:54 EDT 2011


On Mar 27, 2011, at 5:12 PM, Rodrigo Rosenfeld Rosas wrote:

> Em 27-03-2011 06:55, David Chelimsky escreveu:
>> On Mar 26, 2011, at 9:43 PM, Rodrigo Rosenfeld Rosas wrote:
>> 
>>> Hi David and fellows,
>>> 
>>> I know this subject has already been discussed here and there are already some attempts to support the given-when-then-and syntax in Rspec, like the links below:
>>> 
>>> https://gist.github.com/206969
>>> https://github.com/jimweirich/rspec-given
>>> 
>>> First, I don't understand the reason to adopt the capital method names Given/When/Then/And instead of given/when/then/and.
>> 'when', 'then', and 'and' are all Ruby keywords. You can define them as methods, but unpleasant things happen.
> 
> Right! Of course there was a reason! How didn't I think about this before?! :)
> 
>>> Then, I don't think we can call this some real support for that syntax because "example" is not meant to be executed in any special order in Rspec nor it can be defined that any additional "then/and" should be aborted if a prior expectation wasn't met. And both solutions seem to work by simple aliasing "example" and "describe".
>>> 
>>> I never used Cucumber because I find that working with it is cumbersome for most cases.
>>> 
>>> I've been working mostly with Grails on my daily job for about 2 years now since I moved to my current job. I wanted something like Rspec for Groovy and Grails and while searching for some alternative, I found EasyB:
>>> 
>>> http://www.easyb.org/
>>> 
>>> Actually I liked the story syntax they provide as well as the reports and I found it would be useful for Rspec to incorporate that style too:
>>> 
>>> http://www.easyb.org/howtos.html
>>> 
>>> What do you think?
>> Adding first class support for feature/scenario/given/when/then would be a fundamental change to rspec-core, requiring much more than simply exposing new syntax. We'd need new formatters, new ways to define shared code, new command line options, etc, etc. I think this would require too much work, and make the libraries more complicated to use and maintain.
> 
> I really don't intend to replace current Rspec syntax. I like it too. I just proposed to add these new syntax because I feel it is very useful, both for avoiding several expectations by example when you need something sequential and for generating good reports for some use cases.
> 
> I said that because I don't agree that the library would be more complicated to use

Then we disagree.

> although it would certainly add more maintainance work. It would just add some new syntax if the user wants it. For instance, we could do like EasyB does. Suggest creating .spec and .story files. Each one of them would use one of the approaches. That way, it wouldn't make it more complicate to use. But I agree that maintainance would be harder, but on the other hand the user experience could improve a lot for some users.
> 
>> I like the direction that rspec-given takes us: an extension library on top of RSpec. As you point out, it could be made more valuable by adding rich concepts like "steps" that are different from "examples", but I think that could be done within the context of rspec-given (or any other extension), and I'd gladly consider adding new extension points that would make that job easier.
> 
> Right, that's awesome and I agree with you. But as it happens with rspec-rails, I think that rspec-given could be some official module to Rspec to show users that Rspec is commited to support this syntax too.

Except that RSpec is _not_ committed to supporting this syntax.

> The documentation could be integrated as well as it also happens to rspec-rails.
> 
> How do you see this possibility?

I appreciate your enthusiasm for this, and encourage you to write an extension library that provides what you're looking for, but I don't envision this getting merged into RSpec proper for the foreseeable future.

Cheers,
David


More information about the rspec-users mailing list