How to get the list of month names in the current locale ?

[RenjiSann]

Hi !

In my mind, the function I want to do has a very simple signature:

fn get_abbr_month_names(locale: Locale) -> Result<Vec<String>>;

I can't figure out how to do this properly.

The furthest I was able to go right now is making a FixedCalendarTimeNames, but I'm unsure this even is the right way to go

    let mut fmt: FixedCalendarDateTimeNames<Gregorian, fieldsets::M> =
        FixedCalendarDateTimeNames::try_new(locale.into()).unwrap();

Any advice on this matter ?
Thank you !

[sffc]

The problem is that there isn't such a thing as a "list of month names in the current locale" in general. For example, in multiple calendar systems (including Hebrew, Chinese, and Dangi), the list of month names depends on the year, not just the locale.

The code you wrote above will allow you to fetch the Gregorian month names.

[RenjiSann]

Ah, looks like it's specifically the long names that are broken for French with sort but the short ones work. Lol.

From what I can see, it really just tries to match the prefix of the given string to an abbreviated month name. In the french locale, abbreviated names mostly end with a dot, so it won't match the full name. however, in english locale, abbreviated names really just are the 3-4 first characters of the full name, so the full name will match as well.

[Manishearth]

Anyway, the way to get what you want to get is to:

• Use https://docs.rs/icu_datetime/latest/icu_datetime/provider/struct.Baked.html
• Load data for the marker type DatetimeNamesMonthGregorianV1. You will need to do this multiple times to get the different month names sizes, with data marker attributes 3s, 4s, 5s, and 6s (these come from here, the "s" stands for standalone, the numbers are the number of times M is repeated in a CLDR pattern)
• You will get a MonthNames. Gregorian only uses Linear.

These are all unstable APIs, in theory we could add a stable API that provides this functionality but as I mentioned we are reluctant to support locale aware parsing. @sffc might have a writeup somewhere about why.

Note that this may not handle numeric months like in Chinese (1月，2月), since that's constructed from a formatted number and a pattern. That gets trickier.

[sffc]

What @Manishearth said is correct, but you can also get the month names using the public API you suggested in the OP by calling it multiple times, once for a day in each month of a given year.

[sffc]

I am not familiar with the specific ins and outs of coreutils, but where does GNU Utils get its month parsing data? Could you source it from the same place?

[RenjiSann]

I am not familiar with the specific ins and outs of coreutils, but where does GNU Utils get its month parsing data? Could you source it from the same place?

The coreutils are getting their data from the nl_langinfo(3) function available in the glibc, which directly gets data from the locale tables generated with locale-gen

However, in uutils, we decided to use icu for l10n, because apart from making calls to the libc, we have no way to fetch nor to decode the locale data. I personnally failed to find relevant material regarding the format of these files.

[sffc]

Since it sounds like you are trying to use this data for parsing, you may want to read my blog post on the subject:

https://blog.sffc.xyz/post/190943794505/why-you-should-not-parse-localized-strings

Basically, the way to solve the problem is to avoid an architecture that requires you to parse month names, which can be done much of the time. I acknowledge that there are still use cases that need to do this, and I encourage use-case-specific solutions for them.

[RenjiSann]

I agree parsing localized strings is bad. I wouldn't have done it if it were not for the sake of matching the features of GNU coreutils :/

[RenjiSann]

Here's how I eventually did it

pub fn get_abbr_month_names(loc: Locale) -> Result<Vec<String>, DataError> {
    if loc == DEFAULT_LOCALE {
        const DEFAULT_ABBR_MONTHS: &[&str] = &[
            "JAN", "FEB", "MAR", "APR", "MAY", "JUN", "JUL", "AUG", "SEP", "OCT", "NOV", "DEC",
        ];
        return Ok(DEFAULT_ABBR_MONTHS
            .iter()
            .map(ToString::to_string)
            .collect());
    }
    let data_locale = DataLocale::from(loc);

    let request = DataRequest {
        id: DataIdentifierBorrowed::for_marker_attributes_and_locale(ABBR_STANDALONE, &data_locale),
        metadata: DataRequestMetadata::default(),
    };

    let response: DataResponse<DatetimeNamesMonthGregorianV1> =
        icu_datetime::provider::Baked.load(request)?;

    match response.payload.get() {
        MonthNames::Linear(months) => Ok(months.iter().map(ToString::to_string).collect()),
        _ => todo!("unsupported"),
    }
}

[Manishearth]

Yeah, that should work, with the unsupported case probably just falling back to the regular months.

reminder that these APIs are unstable. You're building binaries so it's probably not a huge deal but you may wish to be careful about updating ICU4X versions.

I think in an upcoming release provider is going to be renamed to provider_unstable

[RenjiSann]

I'll be extra careful with the bumps then ^^ thank you :)