How to get a value of numpy ndarray at index or NaN for IndexError?

Question

My question is similar to this one

but I want to use multidimensional numpy array.

Can this be done without checking the indexes beforehand?

data = np.ones((2,3,4))
idx1 = [0,1]
idx2 = [1,2]
idx3 = [0, 100]
data[idx1, idx2, idx3]

Desired output:

array([1., np.nan])

Does my answer solve your problem? If so, can you please mark it as accepted? — Nils Werner, Apr 04 '19 at 10:24

score 0 · Accepted Answer · answered Feb 15 '19 at 10:56

Given our data

import numpy as np    
data = np.ones((2,3,4))
idx = [[0,1], [1,2], [0, 100]]
shape = np.array(data.shape)

You can get any index column that extends past your original array using

invalids = np.any(idx > shape[:, None] - 1, axis=0)

and clip your indices to valid values using

valids = np.clip(idx, 0, shape[:, None] - 1)

these indicies can be used to index our array

out = data[valids.tolist()]

and the mask of invalid indices can be used to set outliers to nan

out[invalids] = np.nan
# array([ 1., nan])

1 Answers1