Use a more correct test for emoji

[robintown]

Type: defect

Fixes element-hq/element-web#20824.

---

Here's what your changelog entry will look like:

🐛 Bug Fixes

• Use a more correct test for emoji (#7685). Fixes Shavian unicode text displays as emoji text (larger font)  element-hq/element-web#20824. Contributed by @robintown.

[t3chguy]

Looks sane, I wonder how negative the performance impact will be though

Maybe it'd be good to do some performance analytics as maybe using the other regex as a shortcircuit to check whether its needed to run the expensive one will be a net positive performance impact.

[trosel]

What is the difference between the two algorithms?

[robintown]

@trosel mightContainEmoji just looks for characters outside of the normal Multilingual Plane, which is fast but results in false positives for things like Shavian characters, while EMOJIBASE_REGEX is much more sophisticated as it has to look for surrogate pairs, variation selectors, zero width joiners etc. to determine if something is actually an emoji.

[trosel]

@robintown would it be possible to keep a fast check but instead look for unicode range numbers?

Emoji is U+1F600..U+1F64F

Shavian is U+10450..U+1047F

[robintown]

@trosel That's not super feasible, since emoji are scattered across many different Unicode blocks, and the same character might be rendered as an emoji or a simple black-and-white textual symbol depending on what variation selectors are present. If we were to create a regex that accounts for all of these factors, we'd basically just be recreating the same regex as Emojibase.