On Sat, Mar 28, 2009 at 07:27:07PM -0600, Idris Samawi Hamid ادريس سماوي حامد wrote:
> On Fri, 27 Mar 2009 06:05:03 -0600, Khaled Hosny <khaledhosny@eglug.org>
> wrote:
> If it's not so complex, have you considered overloading font features with
> fea files? Now that Adobe has (finally!) completed a major overhaul of the
> spec, including the features neeeded for arabic, that's an option.

Yes, it is an option, but not for complex operations as you know, and
also font dependant (not all fonts use Adobe glyph names etc.)

> But for your more complex needs, see the m-translate module. Maybe that
> can be developed further? Note that Hans has pointed out that this is NOT
> a core module and that it is preferable to fix the source beforehand.

That module shows exactly the problem I mentioned in my earlier message,
it processes the whole input file not only the text buffers that will be
actually typesetted, so If I've an \Idris macro it'll get replaced as
well, or I've to parse the file myself which isn't really an option.
OTP fits nicely in this, but it isn't as powerful as lua, and there seem
not to be much interest in maintaining it in the future.

I think, being able to process the to-be-typesetted text in lua is
desirable and have many use cases. Now we have callbacks for input
buffer, which is not very handy when you need to handle only the textual
material without messing with TeX commands and macros, and callbacks for
token and node list, which make very complex what is otherwise simple
task, like 1st -> 1\high{st} conversion, etc.

Regards,
 Khaled


-- 
 Khaled Hosny
 Arabic localiser and member of Arabeyes.org team
 Free font developer