I actually found the comparative metrics more interesting than the article. Anybody know of similar metrics for boost, perl and qt's re performance?

I'm not sure how useful comparing this to other existing regex implementations is. As the article states, this algorithm supports only a small subset of the functionality of the other implementations.

What's most interesting to me is the massive speedup offered by the JIT, as well as the beauty and simplicity of the regex implementation. Giving readable pure Python code a 1200x speedup is amazing. It's remarkable that it's so much faster than the similarly naive Java implementation.

True, I'm not as interested in the specific implementation in the article vs. some of the other established implementations.

There are also some metrics comparing the performance of a couple RE engines to LPEG in the LPEG paper (http://www.inf.puc-rio.br/~roberto/docs/peg.pdf). Spoiler alert: LPEG is usually faster.