Consider using unicode regex in `str_to_snake()` and friends

[arnaudgallou]

The new str_to_snake() and str_to_kebab() rely on non-unicode regex, which may not work in many languages:

stringr::str_to_snake("IlÉtaitUneFois")
#> [1] "ilétait_une_fois"

stringr::str_to_snake("ΜιαΦοράΚιΈνανΚαιρό")
#> [1] "μιαφοράκιένανκαιρό"

stringr::str_to_snake("ЖилиБыли")
#> [1] "жилибыли"

^{Created on 2025-09-24 with reprex v2.1.1}

Could you consider using unicode regex instead?

[hadley]

Would you like to do a PR? Maybe these functions do also need to take a locale argument. One problem, which maybe we can't overcome is illustrated by Dutch, which has a ij digraph which counts as a single letter. I think ideally we'd use some sort of boundary iterator, but we probably don't want to go that deep.

[hadley]

Probably the best compromise is just to switch to [:lower:], [:upper:], and [:digit:].

[arnaudgallou]

Sure I can do a PR in the upcoming days.

According to ChatGPT, ij is the only digraph treated like a single letter in modern languages so it's probably not worth supporting this specific case for now. I can't suggest an elegant solution to support it anyway.

Probably the best compromise is just to switch to [:lower:], [:upper:], and [:digit:].

I don't think there's any benefit of using these pre-built classes over unicode regex. They are actually more limited. For example, there's no pre-built classes to match special characters or symbols that I'm aware of. Special characters are not replaced in the current implementation when they should IMO:

stringr::str_to_snake("foo$bar")
#> [1] "foo$bar"

^{Created on 2025-09-25 with reprex v2.1.1}

Edit:

Just realized we can match special characters with [^[:alnum:]] if you prefer to use pre-built classes.

[hadley]

I'm not sure what you mean by unicode regex.

[arnaudgallou]

I'm talking about these: https://www.regular-expressions.info/unicode.html

[hadley]

As far as I can tell, \p{Lowercase_Letter} is equivalent to [:lower:] with stringi.