Basically, because the macro ‘creates’ code in the macro definition file, there’s no association with a particular line of code in the original test file. This leads to stacktraces (on error) which appear to be completely unrelated to the actual test file, and potentially a lot of digging to figure out where the problem is.

So while I agree that Shoulda (and similar libraries) are very useful for context and the ’should’ testing syntax, I’d warn against the macro functionality as having a good chance of falling into the ‘clever’ trap.

For the ‘inner setup first’, I think what might work is something like ‘acts_like “{other context}”‘, which would run the setup from the named context before running the tests. I _think_ the Context gem has (or had) something like that, but Context has other problems that made it not work well for me, so I fell back to Shoulda, and never explored that too much.

– Morgan