Thirteen languages for dates, weekdays and money

t.format("D MMM YYYY") gives 3 Jun 2024. Under fr_FR it gives 3 juin 2024, and under de_DE, 3 Juni 2024. Month and weekday names shipped in 0.20.0, in thirteen languages, and sys.locales() names what ships.

time.format gained MMM and MMMM for months, ddd and dddd for weekdays, and A and a for the AM/PM marker, which is what h and hh were waiting on. The render layer gained matching components, so a template can reach the same names:

@{d:month.short} @{d:day}, @{d:year}
@{d:weekday}

One rule covers all of them: a formatter names a field, and .short is the only modifier. Both surfaces read the same table, so t.format("MMMM") and @{d:month} cannot disagree about what June is called. A locale SCUA does not carry falls back to English rather than failing.

#Date order now follows the region

Date order follows the region now, not just the language. en_GB, en_AU, en_NZ, en_IE, en_IN and en_ZA are day-first; en_US and a bare en are month-first, as before. Every non-English locale was already right, which is why 6/3/2024 under en_GB looked correct until a reader in the UK took it as the sixth of March.

#money.format now follows the locale

1.234.567,89 €     de_DE
1 234 567,89 €     fr_FR
€1,234,567.89      en_US

Separator, decimal point and symbol placement all follow the run locale. Per-currency minor units are unchanged, so JPY still renders with no decimals. Under the neutral default and en_US the output is byte-identical to 0.19, so if you have never set a locale nothing about your program moved.

Money values can be taken apart, too:

money.amount(m)
money.code(m)

money.of(money.amount(m), money.code(m)) gives back exactly what you started with, at the original scale, including for zero-decimal currencies like JPY and three-decimal ones like KWD. Use those instead of parsing a formatted string, where $ covers USD, CAD and AUD alike and tells you nothing about which one you have.

#What SCUA knows about a locale

SCUA carries a fixed set of locale data, and the reference now lists it both ways, so you can tell at a glance whether you need a package for something.

It has month and weekday names and their abbreviations, the order a date's fields go in, number separators and grouping, and where a currency symbol sits.

It does not have language-correct sorting, ordinals like 1st, numbers written out as words, accounting or engineering number formats, timezone display names like "Pacific Standard Time", or transliteration between scripts. Those are open-ended and large, and they belong in a package or in the host application.

Timezones are a separate thing from the list above, and worth spelling out. --tz takes a numeric offset: --tz=+02:00, --tz=+05:45, or --tz=UTC. IANA zone names like Europe/Berlin are not accepted, so a program that starts from one has to turn it into an offset itself.

The locale how-to covers --locale and --tz in full, including --locale=host --tz=host for picking both up from the machine.