Alex Shinn scripsit:

> I actually need to separate the features better because things like
> "non-greedy" repetitions can actually be supported by non-backtracking
> implementations, "if" is just a shortcut for "or" with a look-ahead,
> and I have to think about whether it's possible for atomic/commit to
> be supported without backtracking.
>
> Ultimately with a little effort everything can be supported.  One
> trick to support backreferences in DFA impls is to replace them with
> .* and use post-processing to verify.  So it's more a matter of what's
> readily available, not what's possible.

In that case I think there should be a unified system and the feature
mechanism should be abandoned.

Editorially, you shouldn't say that backreferences are prohibitively
expensive: if you need them, you need them.  Just drop "prohibitively".

--
You escaped them by the will-death              John Cowan
and the Way of the Black Wheel.                 xxxxxx@ccil.org
I could not.  --Great-Souled Sam                http://www.ccil.org/~cowan