WIP: Removing string evals

[alindeman]

No description provided.

[xaviershay]

I'd expect this to be faster, but did you benchmark it?

[myronmarston]

I'd expect this to be faster, but did you benchmark it?

I certainly don't want to discourage benchmarking (more data is always nice!) but we've been changing all string evals to non-string evals for RSpec 3. I find the non-string form vastly easier to reason about and maintain, and Tenderlove's blog post dispels the common "dynamic methods defined with string evals are faster" advice I used to hear. In other words, I'm generally happy to switch to non-string eval w/o seeing a benchmark but would want to see one if someone wanted to switch to a string eval. YMMV, of course :).

[JonRowe]

This LGTM and I'm also in favour of converting away from string evals so also don't mind no benchmarking. Merge away in my books.

[fables-tales]

We didn't benchmark this. I'm going to merge this as is, I agree with @myronmarston and @JonRowe that string evals are much harder to reason about than real block evals.