Islamic Calendar Types

Tickets

#5525 Disambiguation of Islamic calendar variants

Background

In CLDR, we have two Islamic calendar types available - "islamic" (astronomical?) and "islamic-civil". They originally came from the ICU implementations that seems to be largely influenced by the book - Calendrical Calculations, Nachum Dershowitz, Edward M. Reingold. In this book, "civil" calendar uses the simple arithmetical algorithm based on CE 622 July 16 Friday (Julian calendar). ICU's "islamic" (astronomical) is based on astronomical calculation of moon phase based on a certain location. The intent of latter is to provider better approximation of Hijri calendar actually used by countries, but it's not quite perfect.

Microsoft also provides two calendar types. In .NET, these are named HijriCalendar and UmAlQuraCalendar. The former was introduced a long time ago. According to this page [http://www.staff.science.uu.nl/~gent0113/islam/islam_tabcal.htm], the algorithm used by MS's "HijriCalendar" is nothing more than simple arithmetic one, that is pretty similar to ICU's "islamic-civil" calendar. I compared MS's "HijriCalendar" with ICU's "islamic-civil" calendar side by side and I found ICU's islamic-civil is always one day after the MS's HijriCalendar. My understanding is that MS's implementation just use CE 622 July 15 (Julian) as the epoch date, that is one day before the ICU's implementation.

Note: The link in the above reference page [http://www.staff.science.uu.nl/~gent0113/islam/islam_tabcal_variants.htm] indicates MS's "Kuwaiti Algorithm" is type I (using Intercalary years with 355 days in each 30-year cycle with 2,5,7,10,13,15,18,21,24,26 &29), but HijriCalendar class on .NET 4.5 seems to use type II (2,5,7,10,13,16,18,21,24,26 & 29).

MS's UmAlQuraCalendar supports Umm al-Qura calendar introduced by Saudi Arabia [http://www.staff.science.uu.nl/~gent0113/islam/ummalqura.htm].

Recently, JSR-310 (formerly - Joda Time) was approved and the whole new date/time package will be integrated into Java 8. JSR-310 folks want to identify calendar type using BCP 47 locale extension. So they proposed to add a few islamic calendar types - one of Um al-Qura, one for Saudi Arabia sighting.

Islamic Calendar Variants

Traditionally, the first day of month is the day of the first sighting of the hilal (crescent moon) shortly after sunset. Because of the sighting of the hilal is affected by various factors not predictable (clouds or brightness of the sky, geographic location and others), it is impossible to determine dates by calculation. This traditional practice is still followed in the majority of Muslim countries. In this proposal, we categorize all of sighting based Islamic calendar as a group of religious calendars.

Because the religious calendar date is not determined precisely, there were several algorithms developed for non-religious purposes. A group of Islamic calendar algorithm defined by simple rules without astronomical calculation is called tabular Islamic calendars. Some of them are used for calculating approximate dates and some are used for administrative purpose in some locations. ICU's civil calendar is one of these and Microsoft's "HijriCalendar" is another one in this group.

According to RH van Gent, there are several known variations of tabular calendar schemes [http://www.staff.science.uu.nl/~gent0113/islam/islam_tabcal.htm].

Also, there are two possible epoch dates for each scheme.

• • Epoch type 'a' ("astronomical" or Thursday epoch): CE 622-07-15 (Julian)

  • Epoch type 'c' ("civil" or "Friday" epoch): CE 622 07-16 (Julian)

With his categorization scheme, Microsoft's "HijiriCalendar" is Type IIa (Note: He originally categorized this one as type Ia - but he updated it to type IIa recently). ICU's civil calendar is Type IIc.

There is another algorithmic calendar called Umm Al-Qura. This algorithm is used by Saudi Arabia for administrative purpose. Unlike tabular calendars, the algorithm involves astronomical calculation, but it's still deterministic. Umm Al-Qura is also supported by Islamic communities in North America and Europe. This algorithm is implemented by Microsoft as "UmAlQuraCalendar". There is a request to implement the algorithm in ICU (#8449 Add Um Alqura Hijri Calendar Support).

So Islamic calendar variants are categorized as below:

• • Religious: Based on the sighting of the hilal. Actual dates varies by location.

  • Algorithmic

    • Tabular: ICU "civil", Microsoft "HijriCalendar"

    • Umm Al-Qura: Saudi Arabia and others, Microsoft "UmAlQuraCalendar"

Hierarchical Calendar Type Subtags

CLDR contains a lot of locale data for formatting dates associated with various calendar types. But such calendar algorithm variants are irrelevant to resolving date format symbols and patterns. For example, the same instant may fall on to different dates depending on what Islamic algorithm variant is actually used. However, when formatting code creates a text representation of the dates, symbols (month names, day of week names ...) and patterns are most likely shared by all Islamic calendar variants.

This proposal will introduce a new policy in the calendar type name space.

• • A calendar type might be represented by multiple subtags

  • When a calendar type consists from multiple subtags, corresponding formatting data might be resolved by prefix match

For example, for a given calendar type "ca-xxx-yyy-zzz", if formatting data for type "xxx-yyy-zzz" is not available, them the formatting data for type "xxx-yyy" is used as the fallback, then finally "xxx".

Proposed Calendar Type Value Changes

The table below specifies all of the proposed changes.

Below is the part of the actual XML contents (common/bcp47/calendar.xml)