Certain Emoji cannot be used as character literals

[lyricsboy]

I was going thru Rustlings, and in primitive_types2.rs it encourages you to use an Emoji for the character literal.

    let // Finish this line like the example! What's your favorite character?
    // Try a letter, try a number, try a special character, try a character
    // from a different language than your own, try an emoji!
    if your_character.is_alphabetic() {
        println!("Alphabetical!");
    } else if your_character.is_numeric() {
        println!("Numerical!");
    } else {
        println!("Neither alphabetic nor numeric!");
    }

I tried to use the emoji 1️⃣ (Keycap Digit One) as the character and it made rustc sad:

error: lifetimes cannot start with a number
  --> exercises/primitive_types/primitive_types2.rs:19:26
   |
19 |     let your_character = '1️⃣';// Finish this line like the example! What's your favorite character?
   |                          ^^

error: unknown start of token: \u{20e3}
  --> exercises/primitive_types/primitive_types2.rs:19:29
   |
19 |     let your_character = '1️⃣';// Finish this line like the example! What's your favorite character?
   |                            ^

error[E0762]: unterminated character literal
  --> exercises/primitive_types/primitive_types2.rs:19:30
   |
19 |     let your_character = '1️⃣';// Finish this line like the example! What's your favorite character?
   |                            ^^

The same kind of compiler errors can be produced with the following code:

fn main() {
    println('1️⃣');
}

Rust version: rustc 1.59.0 (9d1b210 2022-02-23)

[thomcc]

Unfortunately some emoji, like the one you've chosen, are actually a sequence of multiple unicode scalars. I'm not 100% sure what the solution here is, but there are hard technical limitations that prevent us from allowing these emoji in character literals.

I do think the diagnostics could use work though, and it might even be worth adjusting the rustlings text somehow, to avoid encouraging students to run into this issue (I'm not sure how, though).

[lyricsboy]

Yeah, it's a tricky one. Interestingly, in VS Code, if I press delete/backspace with my cursor after the troublesome emoji, it looks like it turns into a normal 1 but there is a stealth combining mark still there, which produces a different diagnostic:

error: character literal may only contain one codepoint
 --> emoji_test.rs:2:13
  |
2 |     println('1️');
  |             ^^^
  |
note: this `1` is followed by the combining mark `\u{fe0f}`

[ChrisDenton]

Yeah, the keycap digit one is made up of three Unicode scalar values:

Scalar value	name	meaning
U+0031	Digit one	1
U+FE0F	Variation Selector-16	Changes the appearance of the previous scalar
U+20E3	Combining Enclosing Keycap	Combines to create a keycap emoji

To test this out try using .chars() double quotes instead ":

fn main() {
    println!("1️⃣");
    println!("{:?}", "1️⃣".chars());
}

To be honest a Rust char isn't great for this sort of thing because it's essentially an implementation detail of Unicode.