Sun incorrectly reported to stay below horizon

[Kenneth Henderick]

At my location the run is incorrectly reported to stay below the horizon, something that doesn't happen here:

AstralError Traceback (most recent call last)
<ipython-input-5-b7a1432d534a> in <module>()
----> 1 l.sun(local=True)

/usr/lib/python2.7/site-packages/astral.py in sun(self, date, local)
    707 date = datetime.date.today()
    708
--> 709 sun = self.astral.sun_utc(date, self.latitude, self.longitude)
    710
    711 if local:

/usr/lib/python2.7/site-packages/astral.py in sun_utc(self, date, latitude, longitude)
   1510 """
   1511
-> 1512 dawn = self.dawn_utc(date, latitude, longitude)
   1513 sunrise = self.sunrise_utc(date, latitude, longitude)
   1514 noon = self.solar_noon_utc(date, longitude)

/usr/lib/python2.7/site-packages/astral.py in dawn_utc(self, date, latitude, longitude, depression)
   1547 return self._calc_time(depression, SUN_RISING, date, latitude, longitude)
   1548 except Exception as ex:
-> 1549 raise AstralError(('Sun remains below the horizon on this day, '
   1550 'at this location.'))
   1551

AstralError: Sun remains below the horizon on this day, at this location.

If I patch the _calc_time method to re-raise the exception, I get this:

...
/usr/lib/python2.7/site-packages/astral.py in _calc_time(self, depression, direction, date, latitude, longitude)
   2326 solarDec = self._sun_declination(newt)
   2327
-> 2328 hourangle = self._hour_angle(latitude, solarDec, depression)
   2329 if direction == SUN_SETTING:
   2330 hourangle = -hourangle

/usr/lib/python2.7/site-packages/astral.py in _hour_angle(self, latitude, declination, depression)
   2299 h = (n / d) - t
   2300
-> 2301 HA = acos(h)
   2302 return HA
   2303

ValueError: math domain error

It only happens on solar_depression of 18. Code snippet:

import astral
a = astral.Astral(astral.GoogleGeocoder)
l = a['Ghent,Belgium']
l.solar_depression = 18
l.sun(local=True)

[Simon Kennedy (sffjunkie)]

That there is an error is correct but not for the reason given in the message. It is occurring for a depression of 18 degrees because the sun never reaches that far below the horizon.

[Kenneth Henderick (kenneth-w)]

Aha, interesting. As I'm trying to figure out in which "band" (day, civil twilight, nautical twilight, ...) the sun is, I can most likely figure it out by checking out where exceptions are raised and where not. However, throwing a correct message (or even different exceptions, all inheriting from the same base exception for backwards compatibility) would be easier.