French Revolutionary Calendar

[MenoData]

This will be done in two steps. This issue is only about the first step, namely to introduce this calendar as subclass of CalendarVariant based on the algorithmic proposal of Rommé. Another variant called LEGAL is also possible and will only be applicable on the historic period during the French Revolution.

The real leap year calculation based on autumnal equinox requires astronomical calculations and represents step 2 which will be introduced in a later release as another variant.

Update from 2017-03-24:
Now I have calculations for autumnal equinox available (astro-branch) so the two-step-strategy does not seem to be necessary any longer.

[MenoData]

The sansculottides (complementary days after 12 months at the end of the year) must not be handled as 13th month. Strictly modelling the exact properties of the original historic calendar "as is" is the goal here. Of course, correct naming of months and days is essential, too. In order to realize both goals, an extra hidden element representing either month or sansculottides as TextElement is probably necessary for formatting purposes.

Update from 2017-06-15:

The idea using a hidden text element is not practical because a French Revolutionary date requires two different formats dependent on if it is a sansculottide day or not. And the sansculottide format has only two elements instead of three for normal dates.

[MenoData]

Link: https://en.wikipedia.org/wiki/French_Republican_Calendar
Validity: 1792-09-22/1805-12-31
Validity during commune of Paris:

• 6–23 May 1871 (16 Floréal–3 Prairial An LXXIX)

The implementation should take into account if the dates for the commune of Paris are based on the autumnal equinox method or on an algorithmic variant (not yet clear to me).

[MenoData]

After having done my own equinox calculations: Both the equinox-method and the Rommé-method yield the same dates for the commune of Paris in year 1871. The commune has at least not obviously chosen the Rommé-method although I might be wrong here - possible corrections citing historical sources welcome. This fact indicates favouring the equinox method because of its legal/official status during the French Revolution.

Conclusion: The FrenchRevolutionaryCalendar could also inherit from Calendrical preferring the equinox method. The original idea to inherit from CalendarVariant is based on equal justification of all variants. Support for Rommé in case of super-class Calendrical would still be possible via specialized factory methods, units and getters. Needs some more design thinking...

[MenoData]

Other experiments/implementations about this topic:

ThreeTen/threeten-extra#71 (discussion only about algorithmic details)
https://github.com/caarmen/french-revolutionary-calendar (including some i18n-stuff)
https://github.com/mvanbesien/js-calendars/blob/master/js/republican.js (with day names)

[MenoData]

A good source for the original laws of French Revolution says:

Art. Premier. L’ère des Français compte de la fondation de la république, qui a eu lieu le 22 septembre 1792 de l’ère vulgaire, jour où le soleil est arrivé à l’équinoxe vraie d’automne, en entrant dans le signe de la Balance, à 9heures 18 minutes 30 secondes du matin, pour l’Observatoire de Paris.

The term "équinox vrai" (true equinox) indicates the usage of apparent solar time, not mean solar time, so the equation of time to convert from mean to apparent solar time is obviously necessary, too.

And about the term "franciade":

La période de quatre ans, au bout de laquelle cette addition d’un jour est ordinairement nécessaire, est appelée la franciade, en mémoire de la révolution qui, après quatre ans d’efforts, a conduit la France au gouvernement républicain. La quatrième année de la franciade est appelée sextile.

We need to not overlook the word "ordinairement" (usually), so the franciade is not always four years long as users could interprete it on first glance. Support for franciades in the Time4J-implementation would also be nice.