scipy.interpolate.interp1d与kind =“ previous”不能按预期外推

Question

import numpy as np
from scipy.interpolate import interp1d

a = np.arange(10)
b = a + 100

iterp = interp1d(a, b, kind="previous", bounds_error=False)
print(iterp([-1,-2,5,8,12,25]))
# [nan nan  105.  108. nan nan]

First two nan values makes sense, but last two don't since a previous value is available.

Now if I use extrapolate:

iterp = interp1d(a, b, kind="previous", fill_value="extrapolate")
print(iterp([-1,-2,5,8,12,25]))
# [100. 100. 105. 108. 109. 109.]

The last two 109. makes sense but the first two 100. don't since there is no previous value.

The output I'm looking for is the intuitive [nan nan 105. 108. 109. 109.] . Of course I can take either options above as a base and fiddle with it. But is there a direct way to obtain the desired result?

Answer 1

Ok, this workaround isn't too bad I guess:

iterp = interp1d(a, b, kind="previous", fill_value=(np.nan, b.max()), bounds_error=False)
print(iterp([-1,-2,5,8,12,25]))
# [ nan  nan 105. 108. 109. 109.]