Hyphens in Unicode mode

[brettz9]

With the current behavior of escaping -, there is a problem if used in a Unicode regex.

You can see that:

const esc = require('escape-string-regexp');
new RegExp(esc('-'), 'u');

gives:

Invalid regular expression: /-/: Invalid escape

because this does also:

new RegExp('\\-', 'u');

Would seem to call for a Unicode mode? (The hyphen would still need escaping (and be safe to use) within a character class though.)

I don't know whether you'd want to get into such parsing, but just sharing as it is something I came across.

[sindresorhus]

How would you escape - in Unicode mode?

[brettz9]

Yeah, sorry, I was just thinking, it wouldn't make much sense to escape it at all except in an obscure case like:

new RegExp('[' + esc('unsafe-hyphen') + ']');

...because [ would already be escaped if supplying the whole expression to esc.

If covering all such obscure cases, one would need to escape (or strip) commas too:

new RegExp('(repeatingSequence){' + esc('1,32') + '}');

So I think it should be enough for Unicode mode simply to avoid escaping it at all though of course ensure [ was escaped.

[sindresorhus]

You didn't answer my question. It must be possible to escape - somehow in Unicode mode, right?

[sindresorhus]

As for dropping escaping -, see: e76291d

[brettz9]

There is no need to escape it in Unicode mode (nor is it permissible). The solution is not to escape it (unless within a character class, and then it is the same as escaping a hyphen in non-Unicode mode--with a backslash).

[charmander]

\x2d

[brettz9]

Sure, or \u002d or for Unicode mode, \u{2d}.

I was just speaking to the fact that the hyphen has never needed to be escaped when outside of a character class (though in non-Unicode mode, an extra backslash would just be ignored--as when preceding other non-special characters with a backslash--e.g., /\g/ was ok but /\g/u will err).

But I guess a character code/code point escape does avoid the need for parsing to determine context--I think it should indeed always be safe to use such an escape for -, no matter within a character class or not. So yeah, sounds like a good solution!

[sindresorhus]

Awesome. Let's go with that then.

[brettz9]

Are you ok with it just being added without needing a special "mode" argument? If someone is doing diffs against the results or something unusual, then it would be a breaking change, but I wouldn't think it would impact that many users.

[sindresorhus]

Yes, I'm ok with that. I'll do it as a major release just in case.