fribidi-vs-unicode, PyArabicShaping, PyByeDie, UAX#9 draft, etc

Hi,

Some of you may have noticed that (after about 8 years) I fixed a couple real
bugs (though very insignificant) in the bidi algorithm in FriBidi.  Those were
discovered by running FriBidi against the Unicode BidiTest.txt test suite.
The code for that is here:

  https://github.com/behdad/fribidi-vs-unicode

A while back, I ported the Arabic shaping logic in FriBidi to Python.  That
can be found here:

  https://github.com/behdad/pyarabicshaping

Finally, yesterday I wrote a Python reference bidi implementation from
scratch.  Note, it's a *reference* implementation.  Very slow.  It's 150 times
slower than FriBidi.  It does pass all tests and I'm fairly sure that it's
bugfree now:

  https://github.com/behdad/pybyedie

Finally, there's a new draft of UAX#9 out that adds "bidi isolates".  It's a
major update to the algorithm.  The current draft is at:

  http://www.unicode.org/reports/tr9/tr9-28.html#I1

I'm implementing that in the 6.3 branch of pybyedie.  Eventually I hope it
will make it in FriBidi too, though the N0 rule is hard to implement given
FriBidi's API.

Cheers,
-- 
behdad
http://behdad.org/