Top 100 emojis according to the first click-bait site i found

😂 ❤️ 🤣 👍 😭 🙏 😘 🥰 😍 😊
🎉 😁 💕 🥺 😅 🔥 ☺️ 🤦 ♥️ 🤷
🙄 😆 🤗 😉 🎂 🤔 👏 🙂 😳 🥳
😎 👌 💜 😔 💪 ✨ 💖 👀 😋 😏
😢 👉 💗 😩 💯 🌹 💞 🎈 💙 😃
😡 💐 😜 🙈 🤞 😄 🤤 🙌 🤪 ❣️
😀 💋 💀 👇 💔 😌 💓 🤩 🙃 😬
😱 😴 🤭 😐 🌞 😒 😇 🌸 😈 🎶
✌️ 🎊 🥵 😞 💚 ☀️ 🖤 💰 😚 👑
🎁 💥 🙋 ☹️ 😑 🥴 👈 💩 ✅ 👋

/cc @lyrixx ;)

I started to look at it, and i have some news :)

Almost all concerns the Resources/emoji/build.php script

--

1/ the merge algorithm with the parents locales does not work because there is an intermediary level in the $mapsByLocale (by mb_strlen), so the merge beeing not recursive we lost data there.

foreach ($results as $result) {
    // ...
    $codePointsCount = mb_strlen($emoji);
    $mapsByLocale[$locale][$codePointsCount][$emoji] = $name;
}

// ...

$maps += $mapsByLocale[$parentLocale] ?? [];

--

2/ the arrows can simply be ignored 😃

So something like that should be enough

// 263A FE0F    ; fully-qualified     # ☺️ E0.6 smiling face
preg_match('{^(?<codePoints>[\w ]+) +; [\w-]+ +# (?<emoji>.+) E\d+\.\d+ ?(?<name>.+)$}Uu', $line, $matches);
if (!$matches) {
    throw new \DomainException("Could not parse line: \"$line\".");
}
+if (str_contains($matches['name'], '↑')) {     
+    continue;
+}

--

3/ There was some codepoints "manually" added that messed with the build

// We also add a version without the "Zero Width Joiner"
$codePoints = str_replace('200d ', '', $codePoints);
$emojisCodePoints[$codePoints] = $matches['emoji'];

That created 2 codepoints for the same emoji .. or we need to build a 1:1 map in the end, and that made all the warnings during the build due to false-negative tests.

--

4/ Good news: JSON repository

I played a bit with the script and i discovered there is a CLDR-JSON repository, much easier to manipulate during the build. One locale file could be generated like this (once the data loaded).

foreach (self::getCldrAnnotations() as $locale => $data) {
    $rules = array_map(fn (array $data) => $data['tts'][0] ?? null, $data);
    $localeRules[$locale] = [...$localeRules[$locale] ?? [], ...array_filter($rules)];
}

And seeing how it's complete and simple to parse/reuse, i think we should use it to build the Intl data too

--

5/ I also looked a bit more deeply into the derived annotations (75% of the emoji data weight) ... and if we accept just a "second pass" when we build the transliterator map, we could almost make it seemless.

Something like :

// BEFORE
    '👨🏻‍❤‍💋‍👨🏼' => 'bisou : homme, homme, peau claire et peau moyennement claire',
    '👨🏻‍❤‍💋‍👨🏽' => 'bisou : homme, homme, peau claire et peau légèrement mate',
    '👨🏻‍❤‍💋‍👨🏾' => 'bisou : homme, homme, peau claire et peau mate',
    '👨🏻‍❤‍💋‍👨🏿' => 'bisou : homme, homme, peau claire et peau foncée',
    '👨🏼‍❤‍💋‍👨🏻' => 'bisou : homme, homme, peau moyennement claire et peau claire',
    '👨🏼‍❤‍💋‍👨🏼' => 'bisou : homme, homme et peau moyennement claire',
    '👨🏼‍❤‍💋‍👨🏽' => 'bisou : homme, homme, peau moyennement claire et peau légèrement mate',
    '👨🏼‍❤‍💋‍👨🏾' => 'bisou : homme, homme, peau moyennement claire et peau mate',
    '👨🏼‍❤‍💋‍👨🏿' => 'bisou : homme, homme, peau moyennement claire et peau foncée',

// AFTER
    '👨🏻‍❤‍💋‍👨🏼' => '💋: 👨🏻, 👨🏻, 🏼 et 🏼',
    '👨🏻‍❤‍💋‍👨🏽' => '💋: 👨🏻, 👨🏻, 🏼 et 🏽',
    '👨🏻‍❤‍💋‍👨🏾' => '💋: 👨🏻, 👨🏻, 🏼 et 🏾',
    '👨🏻‍❤‍💋‍👨🏿' => '💋: 👨🏻, 👨🏻, 🏼 et 🏿',
    '👨🏼‍❤‍💋‍👨🏻' => '💋: 👨🏻, 👨🏻, 🏼 et 🏼',
    '👨🏼‍❤‍💋‍👨🏼' => '💋: 👨🏻, 👨🏻 et 🏼',
    '👨🏼‍❤‍💋‍👨🏽' => '💋: 👨🏻, 👨🏻, 🏼 et 🏽',
    '👨🏼‍❤‍💋‍👨🏾' => '💋: 👨🏻, 👨🏻, 🏼 et 🏾',
    '👨🏼‍❤‍💋‍👨🏿' => '💋: 👨🏻, 👨🏻, 🏼 et 🏿',
    '👨🏽‍❤‍💋‍👨🏻' => '💋: 👨🏻, 👨🏻, 🏽 et 🏼',
    '👨🏽‍❤‍💋‍👨🏼' => '💋: 👨🏻, 👨🏻, 🏽 et 🏼',

As they are "modifiers" we may even not need them at all and build that map following the specs !

--

I'll work on it tomorrow (at least the first part, to fix the bug currently impacting ... around half the world population 😅 )

But if someone wants to do all (or some of that) today, feel really, really free to start without me :)