Hi Christian,
Thanks for pointing that out. Turns out it’s the same issue with numpy.int32 as mentioned in this recent post.
It’s been fixed in internal releases, but the latest public release (August 2020) does not support it. (if I recall correctly more recent versions of NumPy necessitated that fix)
So, until the next public release, the workaround is the same as mentioned in that thread (eg, use float, or a NumPy float, instead of a NumPy integer):
lons = np.array([-2, 2, 2, -2,-2], dtype=float)
…or…
lons = np.array([-2, 2, 2, -2,-2], dtype=np.float64)
…or even use float literals in the array…
lons = np.array([-2.0, 2.0, 2.0, -2.0, -2.0])