bug#54970: 28.1; Some emoji no longer display

[Howard Melman <hmelman@gmail.com>]

[-- Attachment #1: Type: text/plain, Size: 4670 bytes --]



> On Apr 17, 2022, at 11:14 AM, Robert Pluim <rpluim@gmail.com> wrote:
> 
>>>>>> On Sun, 17 Apr 2022 10:35:17 -0400, Howard Melman <hmelman@gmail.com> said:
> 
>    Howard> Thanks for all this info.  So on that page, in the second headed section of the 
>    Howard> table "Emoji Font" is where U+1F37D appears.  In the "text-vs" row, which
>    Howard> I think is the case of a lone U+1F37D, I see the emoji glyph in my mac browser.
>    Howard> The description in that header says:
> 
>    Howard>   “text+ts” should be monochrome; everything else should be colorful & monospace.
> 
>    Howard> which matches what I see.  So I think, a lone U+1F37D should be displayed
>    Howard> as an "emoji glyph".  
> 
> _If_ you've specified an emoji font for it, which we donʼt do by
> default, since it has Emoji_Presentation = False,

Ok, so you find out that U+1F37D has Emoji_Presentation = False from
http://www.unicode.org/Public/emoji/1.0//emoji-data.txt
which indicates the default presentation should be text:

1F37D ;	text ;	L2 ;	none ;	w	# V7.0 (🍽) FORK AND KNIFE WITH PLATE

> so you should look at the "Plain" section instead.

Ok, looking in the plain section under text-vs on macos 11.6.5 in Safari
I see an emoji glyph there too for U+1F37D.  This is the plain section 
and U+1F37D is at the end of the top row in this image. There are a lot
of emoji shown in that section but less than in the emojiFont section.



> 
>    Howard> Can emacs be configured to display these lone codepoints via my emoji font?
>    Howard> I gather that's what using the 'symbol script does but also includes more.
>    Howard> Can I (or emacs out-of-the-box) be more selective in the call to 
>    Howard> set-fontset-font or some other api?
> 
> Yes. Try:
> 
> (set-fontset-font t #x1f37d
>    '("Apple Color Emoji" . "iso10646-1") nil 'prepend)
> 
> For a range of codepoints, replace #x1f37d with something like
> '(#x1f37d . #x1f3aa)

Thanks, doing these definitely gets me further to where I'd like:
(set-fontset-font t '(#x1F170 . #x1F6F3) '("Apple Color Emoji" . "iso10646-1") nil 'prepend)
(set-fontset-font t '(#x2139 . #x3299) '("Apple Color Emoji" . "iso10646-1") nil 'prepend)

I'm still confused as to why the above works but this didn't:

(set-fontset-font t 'emoji '("Apple Color Emoji" . "iso10646-1") nil 'prepend)

And I as I look at script-representative-chars, emoji is defined to be  (emoji 127744 128512)
which I think means the hex range x1F300 - x1F600 so shouldn't include x1f37d?
Or does it not because the default expression is text?  And if so how is that
factored into the emoji script symbol passed to set-fontset-font, I don't see
how that's defined other than as this range. And when I specify a range 
directly I get my pretty glyph displayed.

> I donʼt think we should follow what the mac does when it contradicts
> what Unicode is telling us.

I certainly agree with this.   I see that 
https://unicode.org/reports/tr51/#Emoji_Implementation_Notes
says:

	• only fully-qualified emoji zwj sequences should be generated by keyboards and other user input devices.

and working through the definition of fully-qualified emoji
https://unicode.org/reports/tr51/#def_fully_qualified_emoji
a lone U+1F37D is not fully-qualified.  

If I understand emacs' state correctly, insert-char is doing the right thing
because it's just inserting a character.  I think I'm picking an emoji but
I'm not really, I'm picking a single character (in this case U+1F37D). 
A later Emacs will have an emoji input method that would
be like a real emoji picker that lets me insert a proper fully-qualified sequence.

>    Howard> And I'll add, if that's displayed equivalently I'd prefer it, because I wouldn't
>    Howard> have to deal with "extra invisible characters" after the glyph when
>    Howard> using emacs editing commands (unless this is different behavior in 29
>    Howard> than in 28 when I add the variation selector character).
> 
> Those characters get composed, so they get treated as a single
> unit. They really donʼt cause any problems.

Well C-f and C-b seem to move point between them which is somewhat startling.

>>> Modulo `use-default-font-for-symbols'
> 
>    Howard> FWIW this variable set to t for me which I think is the default.
> 
> I meant you should try setting it to 'nil'.

In an emacs -Q in the scratch buffer I inserted a lone U+1F37D 
Toggling use-default-font-for-symbols had no effect on its display.  
Even after I did:

(set-fontset-font t 'emoji '("Apple Color Emoji" . "iso10646-1") nil 'prepend)

Howard

[-- Attachment #2.1: Type: text/html, Size: 8164 bytes --]

[-- Attachment #2.2: Screen Shot 2022-04-17 at 12.01.23 PM.png --]
[-- Type: image/png, Size: 237089 bytes --]