Re: Emacs for Arabic

--- Eli Zaretskii wrote:
> 
> On Tue, 2 Jul 2002, Nadim Shaikli wrote:
> 
> > 2. Regarding the "simplified" Bidi, could I suggest the following
> >    rather simple addition.  Pseudo-code would get the point across
> >    simpler.
> > 
> >      if ( first_char_is_RTL )
> >        Assume entire line is Right-To-Left
> >      else
> >        /* default */
> >        Assume entire line is Left-To-Right
> 
> I cannot say I like this approach.  In my experience, it yields 
> dreadfully wrong results in many practical cases.  (MS-Windows implements 
> something very similar.)

It certainly wouldn't be perfect, but as a simplified hack/approach it
might serve to be useful.  For the various documents that I'm holding on
to, this approach would serve it very well (my perl hacks note that
at least:-).  It could be yet another "Use first-char BIDI algorithm"
next to the current 3 options under the Bidi menu.

> > 3. Your comment about "Handa's available time" pertains to modifying
> >    the code to upload this to emacs' live CVS ?  In other words, this
> >    work/effort will make it to emacs' proper repository, correct ?
> 
> The entire approach to the bidi reordering that Handa-san wrote several 
> years ago was more or less rejected by Gerd Moellmann, who was very 
> unhappy about the way the code caches characters/glyphs before reordering 
> them.  Gerd was convinced that this will slow down the display engine to 
> an unbearable degree.
> 
> That is why Handa-san's code was never added to Emacs, although it exists 
> for quite some time now.
> 
> An algorithm for bidi reordering that fits better into the Emacs display 
> engine was developed, but its integration into Emacs awaits for me to 
> have some free time to work on that.  Unfortunately, I'm working on this 
> alone, since no one has stepped forward willing to hack the Emacs display 
> code.
> 
> >    So that we can expect emacs-21.3 (or emacs-21.4) once released to
> >    have this as native support.
> 
> I very much doubt that.  Emacs 21.3 is a bugfix-only release which is 
> already in pretest; it has no new features.  Emacs 21.4 will be delivered 
> from the current CVS head codebase, but for the reasons I explained 
> above, I don't think it will use the code mentioned here.
> 
> In the meantime, the code mentioned here can be used as a testbed for 
> user-level features and some initial experience with bidi editing.  That 
> will come in handy when bidi support is integrated into Emacs.

Bidi aside, could the Arabic shaping code (ie. the addition of the
'composition-lock-mode') be included ?  Its orthogonal to Bidi and
doesn't encroach.  Granted this might be too late for 21.3, could this
be considered for 21.4 ?  It's __extremely__ useful and a true leap
forward for those of us that have been waiting for it for what seems
like eternity :-)

 - Nadim


__________________________________________________
Do You Yahoo!?
Sign up for SBC Yahoo! Dial - First Month Free
http://sbc.yahoo.com