) mp.set_day(0) # Reset datum to today n = mp.nextphase() # Instantiate generator for _ in range(8): print(next(n))

6.4 DST

Daylight saving time depends on country and geographic location, and there is no built-in MicroPython support. The moonphase module supports DST via an optional user supplied function. DST does not affect the calculation of quarters or phase which is based on the machine clock. If the machine clock runs at a fixed offset to UTC (which is recommended), a DST function can be used to enable reported results to reflect local time.

A DST function takes as input a datetime measured in seconds since the machine epoch (as returned by time.time()) and returns that number adjusted for local time. The following example is for UK time, which adds one hour at 2:00 on the last Sunday in March, reverting to winter time at 2:00 on the last Sunday in October.

def uk_dst(secs_epoch: int):  # Change in March (3) and Oct (10)
    t = time.gmtime(secs_epoch)
    month = t[1]
    mday = t[2]
    wday = t[6]
    winter = secs_epoch
    summer = secs_epoch + 3600  # +1hr
    if month in (1, 2, 11, 12):  # Simple cases: depend only on month
        return winter
    if not month in (3, 10):
        return summer  # +1 hr
    # We are in March or October. Find the day in month of last Sunday.
    ld = (wday + 31 - mday) % 7  # weekday of 31st.
    lsun = 31 - (1 + ld) % 7  # Monthday of last Sunday
    thresh = time.mktime((t[0], month, lsun, 2, 0, 0, 6, 0))  # 2am last Sunday in month
    return summer if ((secs_epoch >= thresh) ^ (month == 10)) else winter

7. Performance and accuracy

7.1 RiSet class

A recalculation is triggered whenever the 24 hour local time window is changed, such as calling .set_day() where the stored date changes. Normally two days of data are calculated, except where the local time is UTC where only one day is required. The time to derive one day's data on RP2040 was 707μs (no twilight calculation, standard clock).

The accuracy of rise and set times was checked against online sources for several geographic locations. The online data had 1 minute resolution and the checked values corresponded with data computed on a platform with 64 bit floating point unit. The loss of precision from using a 32 bit FPU was no more than 3s.

7.2 MoonPhase class

This uses Python's arbitrary precision integers to overcome the limitations of 32-bit floating point units. Results on 32 bit platforms match those on 64-bits to within ~1 minute. Results match those on timeanddate.com within ~3 minutes.