Underscore in replacement string treated as backspace

[hugoduncan]

Using a _ in the replacement string of Regex::replace leads to unexpected behaviour. The _ seems to be treated as a backspace. The documentation should either make mention of this, or this seems to be a bug.

#[test]
fn replacement_with_underscore() {
    let re = regex!(r"(.)(.)");
    let s1 = re.replace("ab","$1-$2");
    let s2 = re.replace("ab","$1_$2");
    assert_eq!("a-b", &s1);
    assert_eq!("a_b", &s2); // Fails here "a_b" != "b"
}

[BurntSushi]

Interesting. It seems that _ is considered part of the capture group name. Since a group with that name doesn't exist, it is replaced with an empty string.

Definitely a bug. Not entirely sure what the right fix is. What is the simplest thing we can do?

[huonw]

What is the simplest thing we can do?

Define that the names of capture groups cannot start with numbers? (i.e. do something similar to Rust identifiers.)

[ekse]

This bug seems to still be there.

let re = Regex::new(r"(.)(.)").unwrap();
let s1 = re.replace("ab","$1-$2");
let s2 = re.replace("ab","$1_$2");
assert_eq!("a-b", &s1);
assert_eq!("a_b", &s2); // Fails here "a_b" != "b"

https://play.rust-lang.org/?version=stable&mode=debug&edition=2018&gist=fa837d651980121ff786eae319965fbc

[BurntSushi]

Ah I see, it looks like I only fixed the first half of this, which was to make capture group names that start with a number illegal. The latter half, which I forgot to do, is to fix expansion such that it treats capture group names starting with a number in the replacement correctly. Thanks for catching this!

[BurntSushi]

Since the docs weren't updated either, I'm marking this as something to be fixed in a regex2 some day. Fixing this now would be a breaking change. We'll just have to live with the wart for the time being.