Rename Islamic calendars to Hijri

[robertbastian]

#6214

• Can we change the era code to be hijri instead of islamic?
  • it could also be ah, given that other calendars use (Latin) abbreviations
• calendrical_calculations also uses "islamic", do we want to change that?

[Manishearth]

Can we change the era code to be hijri instead of islamic?

cc @sffc

Probably not: this matches https://github.com/tc39/proposal-intl-era-monthcode

• calendrical_calculations also uses "islamic", do we want to change that?

shrug it's still the BCP47 name, that's fine for me. Might be worth doing in a future calendrical_calculations breaking update.

[robertbastian]

Our calendars are actually already documented to support only ah: https://unicode-org.github.io/icu4x/rustdoc/icu/calendar/cal/struct.IslamicCivil.html#era-codes

shrug it's still the BCP47 name

Which is irrelevant for the calendrical_calculations crate. Does the book use the term "islamic"?

[Manishearth]

This changes:

• The calendar names
• The AnyCalendar and AnyCalendarKind variants
• The data markers

This does not rename any era codes.

[robertbastian]

it does now!

[robertbastian]

ok fine

[Manishearth]

shrug it's still the BCP47 name

Which is irrelevant for the calendrical_calculations crate. Does the book use the term "islamic"?

It's relevant in that the name is still valid. Yes, the book also calls them islamic. I think the crate matches the book algorithm names most closely.

Our calendars are actually already documented to support only ah: https://unicode-org.github.io/icu4x/rustdoc/icu/calendar/cal/struct.IslamicCivil.html#era-codes

Yeah that's not in line with the tc39 proposal but they keep changing so we can wait.

[sffc]

We should change API names, but not any codes. I don't think this changes codes but I might have missed it.

[robertbastian]

So the docs are wrong? The Hijra calendars all say

This calendar supports a single era code, Anno Hegirae, with code "ah"

but many of them accept multiple codes, and return different codes than "ah". What's the intended behaviour?

[Manishearth]

Yeah, they're inaccurate. The calendars do support a single era (communicating this is the purpose of that line of docs), however that era may have aliases. This stuff has changed a bunch of times.

[robertbastian]

If there's only a singleton era, why validate the era at all?

[sffc]

If there's only a singleton era, why validate the era at all?

My mental model is that a "calendar is a collection of eras", and "an era is a system to project epoch days into an era-year, month-code, day tuple". This framing makes cases like JapaneseModern easy to model: it uses the Gregorian eras until 1868, along with all Gregorian rules.

As another example, I would not be surprised if there was support to make the Hijri calendar use one of the other eras for dates before the Hijri.