At the moment I do it programatically. My steps are named the same as the annotations, so it reads like English. I also include the outcomes, since if I’m running a Scenario before another Scenario and the first Scenario fails, I want to know as soon as possible.

We’re looking at ways of doing it from text but have no real solutions yet. It can’t be hard though, especially if we’re only using one scenario per file – it’s just regexp and file searching.

We’re not doing what you are – running a Background – because I like to explicitly state Givens, and there’s really no difference between running a scenario to buy a fridge, and setting up the state in which I’ve bought a fridge, except that the first exercises the application, which may be simpler than hacking data to set up the state.

I find this statement very interesting as working on the Rspec Story Runners replacement, Cucumber (http://cukes.info/) we have been battling with ways to achieve this (and if we should even allow it).

Our latest move has been to introduce a special scenario referred to as a Background (which should really only contain Givens), which is implicitly run before any of the other scenarios. So we use the background as context for all the other scenarios within a feature/story.

The other solution we use is high level steps to capture common scenario context (so not really using a scenario as a context for another).

Very curious to hear the different ways you achieve using one scenario as the context for another.