GNU bug report logs - #39799
28.0.50; Most emoji sequences don’t render correctly

Previous Next

Full log

View this message in rfc822 format

From: Mike FABIAN <mfabian <at> redhat.com>
To: Eli Zaretskii <eliz <at> gnu.org>
Cc: 39799 <at> debbugs.gnu.org
Subject: bug#39799: 28.0.50; Most emoji sequences don’t render correctly
Date: Fri, 28 Feb 2020 08:36:10 +0100

Eli Zaretskii <eliz <at> gnu.org> さんはかきました:

>> From: Mike FABIAN <mfabian <at> redhat.com>
>> Date: Wed, 26 Feb 2020 15:28:58 +0100
>> 
>> As can be seen in the attached screenshot, some 
>> 
>> 👩‍🦰 U+1F469 U+200D U+1F9B0 woman: red hair
>> 🧑‍🦰 U+1F9D1 U+200D U+1F9B0 person: red hair
>> 
>> don’t render correctly in the screenshot, although they work using the
>> same font (“Joypixels”, version 5.5) elsewhere, e.g. in gedit.
>> 
>> Same result in Emacs when using "Noto Color Emoji", both emoji sequences
>> are rendered as 2 characters each in Emacs
>
> Not 2, 3.  Look more closely, and you will see that the U+200D ZWJ
> character is displayed as a thin (1-pixel) space between the 2 emoji.

Yes.

>> When loading
>> 
>> http://www.unicode.org/Public/emoji/12.0/emoji-zwj-sequences.txt
>> 
>> into Emacs one can see that most sequences don’t render correctly
>> (actually *all* sequences, as far as I can see).
>
> That's just a matter of setting up composition-function-table to
> support these sequences.  For example, try the above again after
> evaluating:
>
>   (set-char-table-range composition-function-table '(#x1F9B0 . #x1F9B3)
> 			(list
> 			 (vector
> 			  "[\U0001F468-\U0001F469]\u200D[\U0001F9B0-\U0001F9B3]"
> 			  2
> 			  'compose-gstring-for-graphic)))

Yes, that does indeed work.

> Patches are welcome to convert the emoji-related files in Unicode's
> character database into appropriate composition-function-table setup,
> similar to the example above.  Some script to be run at Emacs build
> time and produce, say, lisp/emoji.el to populate
> composition-function-table, would be nice (see the Awk scripts in
> admin/unidata as one source of inspiration).

Pango also has a .c file which is generated by a python script from
the Unicode emoji data files to make all these sequences known to Pango.

I can try to write a script. Would it be OK to use Python for such a
script generating emoji.el?

>> Also, when loading
>> 
>> http://www.unicode.org/Public/emoji/12.0/emoji-sequences.txt
>> 
>> into Emacs, one can see that the Flag sequences and skin colour
>> sequences don’t render correctly either (not a font problem, both
>> “Noto Color Emoji” and “Joypixels” support these):
>
> If you mean they are not displayed in correct colors, then Emacs
> doesn't yet support color emoji, we lack some infrastructure for
> that.  Again, work in that area is welcome, it should be relatively
> easy since we now have HarfBuzz support for text shaping.

Actually the color display works already. I tested with current master
(build with cairo) and the emoji display just fine in color.

-- 
Mike FABIAN <mfabian <at> redhat.com>

Previous Next

GNU bug tracking system
Copyright (C) 1999 Darren O. Benham, 1997,2003 nCipher Corporation Ltd, 1994-97 Ian Jackson.