Now might be a good time to figure out off-the-shelf calendar sets in DateTimeFormatter about icu4x HOT 7 OPEN

I would keep all "Islamic"s. I know for a fact that Saudi Arabia uses islamic-umalqura, that Iran uses islamic observational, and that some Arab countries use the tabular and civil calendars. All four are needed.

from icu4x.

I think this is a good idea and is a great example of why the neo model works much better overall.

We may need to eventually figure out what counts as extended but for now "Japanext + community calendars" seems fine.

from icu4x.

• @robertbastian - We could also just exclude JapaneseExtended from AnyCalendar, and if clients want to use it, they need to use TypedDateTimeFormatter.
• @sffc - It would be a pain if a client did need JapaneseExtended but had to use a special code path for it. But, we can cross that bridge later. We can add try_new_extended if needed.

Conclusion: In 2.0, remove JapaneseExtended from AnyCalendar and DateTimeFormatter bounds

LGTM: @sffc @robertbastian

@Manishearth thoughts?

from icu4x.

I'm fine with this. Should we also consider removing any of the Islamic ones? I'm not sure which ones are "major" or in active civil use.

from icu4x.

The CLDR data has each calendar except Julian (and Japanese Extended) listed in at least one locale:

    <calendarPreferenceData>
        <calendarPreference territories="001" ordering="gregorian"/>
        <calendarPreference territories="BD DJ DZ EH ER ID IQ JO KM LB LY MA MR MY NE OM PK PS SD SY TD TN YE" ordering="gregorian islamic islamic-civil islamic-tbla"/>
        <calendarPreference territories="AL AZ MV TJ TM TR UZ XK" ordering="gregorian islamic-civil islamic-tbla"/>
        <calendarPreference territories="AE BH KW QA" ordering="gregorian islamic-umalqura islamic islamic-civil islamic-tbla"/>
        <calendarPreference territories="AF IR" ordering="persian gregorian islamic islamic-civil islamic-tbla"/>
        <calendarPreference territories="CN CX HK MO SG" ordering="gregorian chinese"/>
        <calendarPreference territories="EG" ordering="gregorian coptic islamic islamic-civil islamic-tbla"/>
        <calendarPreference territories="ET" ordering="gregorian ethiopic"/>
        <calendarPreference territories="IL" ordering="gregorian hebrew islamic islamic-civil islamic-tbla"/>
        <calendarPreference territories="IN" ordering="gregorian indian"/>
        <calendarPreference territories="JP" ordering="gregorian japanese"/>
        <calendarPreference territories="KR" ordering="gregorian dangi"/>
        <calendarPreference territories="SA" ordering="islamic-umalqura gregorian islamic islamic-rgsa"/>
        <calendarPreference territories="TH" ordering="buddhist gregorian"/>
        <calendarPreference territories="TW" ordering="gregorian roc chinese"/>
    </calendarPreferenceData>

That said, I don't have a very great sense of which Hijri calendars are needed in which contexts. For example, would it be acceptable if we included islamic-umalqura in the default set but not islamic (observational)? I was having a lot of trouble getting islamic to behave correctly in the stress testing in #4904.

from icu4x.

CC @roozbehp @younies for potential feedback on calendar selection.

from icu4x.

Yeah that's what I'm trying to figure out; it's unclear which ones are actually in use.

from icu4x.