Whow, huge!
This should replace perl's regex engine, as it would blow it away with its builtin unicode features and much saner coding. No longjmp's.
The description in the documentation sounds interesting but I can not find any license, and although that might seem pedantic I can not even read the source code without it.
There's a MIT license in both source files, but I added a copy of it to the project root for those that have this constraint. Thanks!