lark

lark copied to clipboard

Fixes #76.

This is a first attempt.

You can install interegular via pip: pip install interegular. If you want, you can also take a look at the interegular source code.

A few notes:

1. Basically, this is the real functional code (which resides in interregular):

        for a, b in combinations(keys, 2):
            if not self.isdisjoint(a, b):
                yield a, b

I see you already cache the FSMs, so all that's required to prevent duplicate work is to call memoized_disjoint instead.

2. Why does lark have to call mark and is_marked? Seems unnecessary. If you want to avoid duplicate warnings, just keep a set.

3. Why compare regexps (expensive) and only afterwards check a.priority == b.priority? This should be the first test, otherwise comparing them with interregular is pointless.

4. skip_validation means you never use the comparator. So why even create it in the first place?

P.S. re point 3, you can do something like classify(regexps, lambda r: r.priority) to get all the subgroups that should be tested together.

Well, this is now a messed up git history.