How to replace empty values of an array with NaN

Question

I want to create an array with shape: (N,3). But, when it is not possible, I will like to replace missing value with NaN .

Here is my code:

from scipy import spatial
import numpy as np
vertices = numpy.array([[ 0.82667452,  0.89591247,  0.91638623],
                        [ 0.10045271,  0.50575086,  0.73920507],
                        [ 0.06341482,  0.17413744,  0.6316301 ],
                        [ 0.75613029,  0.82585983,  0.10012549],
                        [ 0.45498342,  0.5636221 ,  0.10940527],
                        [ 0.46079863,  0.54088544,  0.1519899 ],
                        [ 0.61961934,  0.78550213,  0.43406491],
                        [ 0.12654252,  0.7514213 ,  0.18265301],
                        [ 0.94441365,  0.00428673,  0.46893573],
                        [ 0.79083297,  0.70198129,  0.75670947]])
tree = spatial.cKDTree(vertices)
iof = tree.query_ball_point(vertices,0.3,n_jobs=-1,return_sorted=False)
faces_eampty = np.empty((len(vertices),3))
faces_eampty[:] = np.NaN
faces = np.array([l[0:3] for l in iof])
faces_eampty[:faces.shape[0], :faces.shape[1]] = faces
np.savetxt("normaltest_2.txt",faces,fmt='%s')

I want the result to be something like this:

faces: [[0 6 9]
 [1 2 4]
 [1 2 NaN]
 [3 4 5]
 [1 NaN NaN]
 [1 2 3]
 [0 1 3]
 [1 2 NaN]
 [5 NaN NaN]
 [0 1 3]]

How can I do this?

Answer 1

Given that you may have 1, 2, or more results from your query for each point, it's best to not turn to an array until the values all have the same shape (required for a numpy array).

from scipy import spatial
import numpy as np
vertices = np.array([[ 0.82667452,  0.89591247,  0.91638623],
                        [ 0.10045271,  0.50575086,  0.73920507],
                        [ 0.06341482,  0.17413744,  0.6316301 ],
                        [ 0.75613029,  0.82585983,  0.10012549],
                        [ 0.45498342,  0.5636221 ,  0.10940527],
                        [ 0.46079863,  0.54088544,  0.1519899 ],
                        [ 0.61961934,  0.78550213,  0.43406491],
                        [ 0.12654252,  0.7514213 ,  0.18265301],
                        [ 0.94441365,  0.00428673,  0.46893573],
                        [ 0.79083297,  0.70198129,  0.75670947]])
tree = spatial.cKDTree(vertices)
iof = tree.query_ball_point(vertices,0.3,n_jobs=-1,return_sorted=False)
faces = [l[0:3] for l in iof]

# append a NaN value if there are less than n values
# where n are the most results from the query from 1 point
_max = max([len(f) for f in faces])

for f in faces:
    if len(f) < _max:
        f.append(np.NaN)

np.array(faces)
>>> array([[ 0.,  9.],
           [ 1., nan],
           [ 2., nan],
           [ 3., nan],
           [ 4.,  5.],
           [ 4.,  5.],
           [ 6., nan],
           [ 7., nan],
           [ 8., nan],
           [ 0.,  9.]])