rust's regex crate versus twitter-text

main.rs

//Rust code to compare the regex crate's version of these unicode categories
//to Ruby/Java/Objective-C's versions as collected by Twitter
//https://github.com/twitter/twitter-text/tree/master/unicode_regex

extern crate regex;

fn main() {
    let mut decimal = Vec::new();
    let mut marks_letters = Vec::new();

    let decimal_pat = regex::Regex::new(r"\p{Nd}").unwrap();
    let marks_letters_pat = regex::Regex::new(r"[\p{L}\p{M}]").unwrap();

    let mut temp = String::with_capacity(4);

    for x in (0u32..0xD800).chain(0xE000u32..0x110000) {
        temp.push(std::char::from_u32(x).unwrap());

        if decimal_pat.is_match(&temp) {
            decimal.push(x);
        }

        if marks_letters_pat.is_match(&temp) {
            marks_letters.push(x);
        }

        temp.clear();
    }

    //prints 550 codepoints versus objc's 540
    //rendered version available at https://shiva.icesoldier.me/rust_twitter_text/decimal_numbers_rust.txt
    print_matches(&decimal, decimal_pat.as_str());
    //prints 111554 codepoints versus objc's 104623
    //rendered version available at https://shiva.icesoldier.me/rust_twitter_text/letters_and_marks_rust.txt
    print_matches(&marks_letters, marks_letters_pat.as_str());
}

fn print_matches(input: &[u32], pattern: &str) {
    println!("# {} code points matched for {}", input.len(), pattern);

    for x in input {
        println!("{}", x);
    }
}