Re: Arabic-Emacs questions

[TAKAHASHI Naoto <ntakahas at m17n dot org>]

Thank you very much for your comments.

Nadim Shaikli writes:

> Is the 'composition-lock-mode' (ie. Arabic shaping; even without
> bidi) being slated to be included to emacs' proper source tree ?

As far as I know, not decided yet.  However, since the current version
of comp-lock is far from perfect, it requires improvements to be
included in the main tree.

Nonetheless, a glyph shaping mechanism like comp-lock is useful not
only for Arabic but also for many Indian scripts.  I believe the
forthcoming Unicode-based Emacs will provide something like comp-lock
to handle those scripts.

> But if I have say DAL (0x062f) followed by any of the tanween
> characters (FATHATAN .. SUKUN -- 0x064b .. 0x0652) the cursor upon a
> search ends up being placed in such a manner that it shifts to the
> left the tanween characters (which looks rather odd).  It would be
> more natural to highlight and to move the cursor in such a manner to
> include the tanween unless a person included a tanween him/herself
> in the search to exclude others.

I do not understand.  I created a file that contains four lines of

> DAL-FATHATAN-NOON
> DAL-MEEM
> DAL-NOON
> DAL-KASRA-NOON

and hit C-s ] ("]" for DAL).  Then

1. The DAL-FATHATAN composion is temporarily released.
2. The first DAL is highlighted.
3. The DAL in the second line, one in the third line, and the
   DAL-KASRA composition in the fourth line are highlighted with a
   different colour.
4. The cursor is placed on the temporarily released FATHATAN.
5. When I hit C-s repeatedly to move the corsor to the fourth line,
   the DAL-KASRA composition is temporarily released as in 1. above.

These are expected hehaviour, arn't they?

> so if I do a search for DAL-NOON, emacs should highlight and move
> the cursor (when the cursor is moved to that character-combo, the
> cursor ought to sit past the "NOON") for all of a,c,d.

Are you saying that incremental-search should behave as if tanweens do
not exist in the buffer?

> Along the same lines, when doing a search only for a tanween
> character (0x064b - 0x0652), emacs doesn't do any highlighting
> although it works properly.

Thank you for pointing it out.  This is a bug we have not noticed.  It
may take some time to fix it.

> Don't know what would seem natural here, but some highlight would be
> nice.

Perhaps a temporary release of the composition as I described above is
acceptable.

> This might generate some debate, but I thought I'd mention it.  It
> "might" be useful to have search ignore repeated TATWEEL (0x0640)
> unless the user specifies it.

Again, are you saying that TATWEEL should be invisible (like tanweens)
for incremental-search unless the user explicitly specifies?

> Any comments on the bidi addition (pseudo code) noted earlier to
> base line direction according to the first character in the line ?

As I wrote before, the current Emacs cannot change the screen
orientation ("base line direction", in your term) in a window.  Thus
your idea is technically impossible to implement.

-- 
TAKAHASHI Naoto
ntakahas at m17n dot org
http://www.m17n.org/ntakahas/