numpy interp decreasing xp

Question

I want to interpolate a numpy array, and the np.interp does almost exactly what I want:

Click to copy

interp(x, xp, fp, left=None, right=None)
One-dimensional linear interpolation.

Except for this bit:

Does not check that the x-coordinate sequence xp is increasing. If xp is not increasing, the results are nonsense.

My xp is decreasing, so which is better: Reversing the direction of both xp an fp:

Click to copy

np.interp(x, xp[::-1], fp[::-1])

or inverting x and xp:

Click to copy

np.interp(-x, -xp, fp)

Or is there an even better way?

Answer 1

Thanks for all those that have given their input, especially @Jaime.

I have experimented a bit, and came to this conclusion:

1) Apart from rounding errors, both methods mentioned by me have the same result.

2) They both take very much the same amount of time

3) I tried the scipy version, but it would reject the assume_sorted flag. Maybe my version of scipy is outdated. I suspect that if that flag is raised, scipy internally sorts the arrays. But the values are sorted, just in the opposite direction, so it doesn't need to do this.

Anyway, I'll use the reverse-direction method:

Click to copy

np.interp(x, xp[::-1], fp[::-1])

Just remember that in this case, you'll also have to reverse left and right, if you need them.

Answer 2

If you have access to Scipy, you could use the function interp1d, which has the keyword assume_sorted=False to handle decreasing arrays.

Edit: This solution handles both the cases of ordered and non ordered x values.

Click to copy

import numpy as np
from scipy import interpolate
import matplotlib.pyplot as plt

# Decreasing array
x = np.arange(0, 10)[::-1]
y = np.exp(-x/3.0)

# Interpolation object
f = interpolate.interp1d(x, y, assume_sorted = False)

xnew = np.arange(3,5)
ynew = f(xnew)   # use interpolation function returned by `interp1d`
plt.plot(x, y, 'o', xnew, ynew, '-')
plt.show()