The matching results of strings containing Chinese characters are incorrect about regexp2

[dlclark]

After more digging I've figure out the difference. In the regexp package \b is only a word boundary for simple ascii [A-Za-z0-9_] so the initial Chinese characters are not considered a word boundary. Since regexp2 (and Rust, Python, and C#) uses full unicode definition for "letter" to define \b the initial Chinese characters are considered a word boundary.

If you expand \b to the full definition ((?<!\w)(?=\w)|(?<=\w)(?!\w)) and then expand the \w to the simplified regexp package definition A-Za-z0-9_ then you end up with this regex which will make regexp2 capture similarly to regexp:

((?<![A-Za-z0-9_])(?=[A-Za-z0-9_])|(?<=[A-Za-z0-9_])(?![A-Za-z0-9_]))(((([*+\-=?^_{|}~\w])|([*+\-=?^_{|}~\w][*+\-=?^_{|}~\.\w]{0,}[*+\-=?^_{|}~\w]))[@]\w+([-.]\w+)*\.[A-Za-z]{2,8}))((?<![A-Za-z0-9_])(?=[A-Za-z0-9_])|(?<=[A-Za-z0-9_])(?![A-Za-z0-9_]))

FYI, since the above regex uses look-arounds it's not compatible with the regexp package.

[dlclark]

The target for this regex engine is to match .NET. In this example https://regex101.com/r/mFmrlg/1 it appears regexp2 matches C#'s behavior (and Rust and Python).

There's an RE2 option that increases compatibility with RE2 (regexp package) but it doesn't work in this case.

I'll spend a bit of time later today to see if I can understand why the difference exists, but since this is such a long regex I may not be able to figure it out. If you have a shorter regex example of the issue that'd help speed up understanding the root difference.