Since 1.37 events missing in twilight calculation

[mworion]

@brandon-rhodes Since I moved to 1.37 I observed a change in my app when calculating the twilight data from:

to:

I calculate the data with

f = almanac.dark_twilight_day(self.app.ephemeris, location)

When looking to the data I see oscillating value for event codes (second row):

2459381.0 3 14.116666666666667
2459381.0 4 15.833333333333334
2459382.0 3 10.649999999999999
2459382.0 4 15.816666666666666
2459383.0 3 10.649999999999999
2459383.0 2 12.4
2459383.0 3 14.066666666666666
2459383.0 4 15.816666666666666
2459384.0 3 10.666666666666664
2459384.0 2 12.433333333333334
2459384.0 3 14.05
2459384.0 4 15.816666666666666
2459385.0 3 10.666666666666664
2459385.0 4 15.816666666666666

From my point this is impossible. It happened from 1.36 to 1.37.
Michel

[brandon-rhodes]

1. What Skyfield version were you previously on — can I look for a difference only between 1.36 and 1.37, or did you upgrade from several versions ago?
2. Is there any chance you could supply a tiny script producing those numbers, so I have a test case to work with? I don't need to plotting code, if that makes it simpler to produce a test case. Thanks!

[mworion]

I switched from 1.36 to 1.37. And I could reproduce this situation just by switching the skyfield version.

    from skyfield.api import Topos, load
    from skyfield import almanac

    location = Topos(longitude_degrees=11,
                     latitude_degrees=47,
                     elevation_m=500)
    ephemeris = load('tests/testData/de421.bsp')
    ts = load.timescale(builtin=True)
    t0 = ts.tt_jd(2459360)
    t1 = ts.tt_jd(2459360 + 30)
    f = almanac.dark_twilight_day(ephemeris, location)
    times, events = almanac.find_discrete(t0, t1, f)

    with open('compare1.36.txt', 'w+') as f:
        for t, e in zip(times, events):
            f.writelines(f'{t.tt:10.4f}, {e}\n')

I see in a diff that th "night" section so event 0 is in 1.36 and misses in 1.37. So to correct it's not oscillating, but missing event. I wrote the content in a file and made a diff. Michel

compare1.36.txt compare1.37.txt

[brandon-rhodes]

Thanks very much for the example script! I've tried it out and done some checking. Try this in your script and let me know if it has an effect:

f.step_days = 0.04

[mworion]

Yes, this sorts it out in my app. Many thanks ! Plot is again OK with 1.37. Added this in my code. Michel

[brandon-rhodes]

For anyone reading this later: the fix went out in Skyfield 1.38.