How to get the connected forms of Syriac letters?

How to get the connected forms of Syriac letters?

[Fabian Rieger]

Is there a method to make Emacs show the connected (initial, medial, 
final) forms of Syriac letters?
I'm using the Unicode-OpenType Meltho fonts, in Iceweasel/Firefox this 
works fine, but Emacs gives nothing but the isolated forms.
    The shaping of Arabic seems to work fine (I can't read it, but the 
letters change to isolated forms and back when a space is added or 
deleted inside a copyed word).
    My fonts:

       (set-default-font "DejaVu Sans Mono 20")
       (set-fontset-font "fontset-default" 'syriac
                 "Serto Jerusalem 30")
       (set-fontset-font "fontset-default" 'arabic
                 "Scheherazade 24")

I tried with Debian Wheezy + Emacs 24.4.50.1 (from londo.ganneff.de) & 
25.0.50.1 built with configure option --with-m17n-flt 
(/usr/share/m17n/SYRC-OTF.flt looks promising but doesn't seem to help) 
following
    www.emacswiki.org/emacs-fr/EmacsSnapshotAndDebian
The problem already was there with win2k & Emacs 24.3 a year ago ...

Thanks

Fabian

Re: How to get the connected forms of Syriac letters?

[Benjamin Riefenstahl]

Hi Fabian,

Fabian Rieger writes:
> Is there a method to make Emacs show the connected (initial, medial,
> final) forms of Syriac letters?

I use something like this in my .emacs with Emacs 24.4.1 on Debian
Jessie:

    ;; Tell Emacs to actually use shaping for Syriac (via m17n).
    ;; \u0700-\u074F: Syriac
    ;; \u0640: TATWEEL
    ;; \u200D: ZWJ
    (set-char-table-range
     composition-function-table
     '(?\u0700 . ?\u074F)
     (list ["[\u0700-\u074F\u0640\u200D]+" 0 font-shape-gstring]))

    (defun benny-setup-fonts (frame)
      (set-fontset-font t (?\u0700 . ?\u074F) "Serto Mardin 16" frame))

    ;; Run font setup whenever a new frame is constructed.
    (add-to-list 'after-make-frame-functions 'benny-setup-fonts)

    ;; Run font setup now if we already have a frame.
    (let ((frame (selected-frame)))
      (if frame (benny-setup-fonts frame)))

benny

Re: How to get the connected forms of Syriac letters?

[Eli Zaretskii]

> From: Benjamin Riefenstahl <b.riefenstahl@turtle-trading.net>
> Date: Mon, 15 Dec 2014 23:32:53 +0100
> 
>     ;; Tell Emacs to actually use shaping for Syriac (via m17n).
>     ;; \u0700-\u074F: Syriac
>     ;; \u0640: TATWEEL
>     ;; \u200D: ZWJ
>     (set-char-table-range
>      composition-function-table
>      '(?\u0700 . ?\u074F)
>      (list ["[\u0700-\u074F\u0640\u200D]+" 0 font-shape-gstring]))

Yes, this is missing from the basic Emacs setup, I will add it
shortly.

But why did you need to mention TATWEEL and ZWJ there?  (And if ZWJ is
needed, then why not ZWNJ as well?)

Re: How to get the connected forms of Syriac letters?

[Benjamin Riefenstahl]

Eli Zaretskii writes:
> But why did you need to mention TATWEEL and ZWJ there?  (And if ZWJ is
> needed, then why not ZWNJ as well?)

Continuing the experiment, I tried ZWNJ (\u200C).  I get a control
character for this similar to the default ZWJ, but when I add it to the
set-char-table-range expressions it works.  So you are right, we should
add that too.

Re: How to get the connected forms of Syriac letters?

[Eli Zaretskii]

> From: Benjamin Riefenstahl <b.riefenstahl@turtle-trading.net>
> Date: Tue, 16 Dec 2014 20:32:40 +0100
> 
> Eli Zaretskii writes:
> > But why did you need to mention TATWEEL and ZWJ there?  (And if ZWJ is
> > needed, then why not ZWNJ as well?)
> 
> Continuing the experiment, I tried ZWNJ (\u200C).  I get a control
> character for this similar to the default ZWJ, but when I add it to the
> set-char-table-range expressions it works.  So you are right, we should
> add that too.

But how is this related to the Syriac script?

Re: How to get the connected forms of Syriac letters?

[Benjamin Riefenstahl]

Eli Zaretskii writes:
>> > But why did you need to mention TATWEEL and ZWJ there?  (And if ZWJ is
>> > needed, then why not ZWNJ as well?)

> But how is this related to the Syriac script?

Ah, I'm sorry, I thought that was understood.

Syriac (and Mandaic) are similar to Arabic script in type, i.e. cursive
scripts where characters adapt according to context ("shaping").  They
re-use the TATWEEL for justification, and also ZWJ and ZWNJ to force or
prevent shaping.

See <http://en.wikipedia.org/wiki/Syriac_script> and
<http://en.wikipedia.org/wiki/Mandaic_script>.

Syriac script is supported by fonts, by libm17n and by Uniscribe.
Support for Mandaic script is just starting, although it has been in
Unicode for some years.

Re: How to get the connected forms of Syriac Letters?

[Fabian Rieger]

Benjamin Riefenstahl wrote
 > I use something like this in my .emacs with Emacs 24.4.1 on Debian
 > Jessie:
 >
 >     ;; Tell Emacs to actually use shaping for Syriac (via m17n).
 >     ;; \u0700-\u074F: Syriac
 >     ;; \u0640: TATWEEL
 >     ;; \u200D: ZWJ
 >     (set-char-table-range
 >      composition-function-table
 >      '(?\u0700 . ?\u074F)
 >      (list ["[\u0700-\u074F\u0640\u200D]+" 0 font-shape-gstring]))
 >
 >     (defun benny-setup-fonts (frame)
 >       (set-fontset-font t (?\u0700 . ?\u074F) "Serto Mardin 16" frame))
 >
 >     ;; Run font setup whenever a new frame is constructed.
 >     (add-to-list 'after-make-frame-functions 'benny-setup-fonts)
 >
 >     ;; Run font setup now if we already have a frame.
 >     (let ((frame (selected-frame)))
 >       (if frame (benny-setup-fonts frame)))

Hi Benny,

that solved the problem. I just had to replace

    (set-fontset-font t (?\u0700 . ?\u074F) ...

with

    (set-fontset-font t (cons (decode-char 'ucs ?\x0700)
  			     (decode-char 'ucs ?\x074F) ...

(The reason for this maybe I'll try to understand next week ...)
and now Syriac really looks fine.

Thanks for your help.

Fabian

Re: How to get the connected forms of Syriac Letters?

[Stefan Monnier]

>    (set-fontset-font t (?\u0700 . ?\u074F) ...
                        ^^
                        '
-- Stefan

Re: How to get the connected forms of Syriac Letters?

[Benjamin Riefenstahl]

Stefan Monnier writes:
>>    (set-fontset-font t (?\u0700 . ?\u074F) ...
>                         ^^
>                         '

Indeed, sorry.