Pyfribidi

* use greenlet's my_build_ext for a better inplace build

* refactor pyfribidi.c module

pyfribidi.c is now compiled as _pyfribidi. This module only handles
unicode internally and doesn't use the fribidi_utf8_to_unicode
function (which can't handle 4 byte utf-8 sequences). This fixes the
buffer overflow in issue 2.

The code is now also much simpler: pyfribidi.c is down from 280 to 130
lines of code.

We now ship a pure python pyfribidi that handles the case when
non-unicode strings are passed in.

We now also adapt the size of the output string if clean=True is
passed.

* Expose reordernsm flag (reorder Non-Spacing-Marks) in log2vis

-- volker <volker.haasbrainbot.com> Thu, Aug 11 12:23:37 CEST 2011

* Change the package number in the setup.py script

-- kobi <kzamirwalla.co.il> Wed, 3 Feb 2009 19:03:11 +0200

* Use fribidi 0.19 and above.
* Re-Change the package name from pyfribidi2 to pyfribidi

-- kobi <kzamirwalla.co.il> Wed, 3 Feb 2009 12:59:24 +0200

* Yoann Roman: Fix Buffer too small with Arabic joining.
* Use fribidi2, fribidi 0.19 and above.
* Change the package name from pyfribidi to pyfribidi2

-- kobi <kzamirwalla.co.il> Fri, 13 Mar 2009 12:59:24 +0200

API Changes:

* Nir Soffer: Accept either unicode or encoded string or subclasses
* Nir Soffer: Accept encoding keyword argument, support any encoding Python
support (encoding it ignored if you send unicode input)
* Nir Soffer: Accept base_direction keyword argument (was only positional before)
* Nir Soffer: Return the same type of input you sent, either unicode or str
* Nir Soffer: Updated docstrings

Internal Changes:

* Nir Soffer: new function log2vis_encoded_string handles strings using either
log2vis_utf8 or log2vis_unicode
* Nir Soffer: Updated docs
* Nir Soffer: Added more tests, organized tests by types

-- kobi <kzamirwalla.co.il> Thu, 18 Jan 2006 00:59:24 +0200