Emoji group matches digits

[Gegy]

Using regex 1.3.9, matching with the \p{Emoji} group yields matches on digits as well as emoji. As far as I could tell, this isn't intentional.

Example to reproduce issue:

use regex::Regex;

fn main() {
    let match_emoji = Regex::new(r#"\p{Emoji}"#).unwrap();

    let matches = match_emoji.find_iter("hello! 😀 😁 1234")
        .map(|m| m.as_str())
        .collect::<Vec<_>>();

    println!("{:?}", matches);
}

This prints ["😀", "😁", "1", "2", "3", "4"], where I'd expect it to print just ["😀", "😁"]

Not totally sure if I'm misinterpreted the intended use of the Emoji group. For now using the pattern [\p{Emoji}--\p{Digit}] is a functioning workaround. 🙂

[BurntSushi]

Using regex 1.3.9, matching with the \p{Emoji} group yields matches on digits as well as emoji. As far as I could tell, this isn't intentional.

It is correct. The is a difference between what an emoji is, what Unicode defines as an emoji and what \p{Emoji} is. An emoji is the thing that a human would recognize as an emoticon in text. Unicode defines emoji via UAX#51 using a set of rules that is assuredly imperfect, but does well in practice. Finally, \p{Emoji} is merely a character class that matches precisely one codepoint among a set of them, where that set is defined by Unicode via the UCD. The purpose of \p{Emoji} is as a convenient tool for implementing the more precise definition described in UAX#51. In that light, digits are indeed defined to be a part of this set, since they can combine with other codepoints to form emoji. This should reveal that simply using \p{Emoji} to recognize emoji is quite problematic because it doesn't implement UAX#51 all on its own. Trivially, it will only match a single codepoint and many UAX#51 emoji consist of several emoji.

I wrote more about this here, when someone asked a similar question about ripgrep, which uses the regex crate: BurntSushi/ripgrep#1623 (comment)