Index into NumPy array ignoring NaNs in the indexing array

Question

I have an array of zeros

arr = np.zeros([5,5])
array([[ 0.,  0.,  0.,  0.,  0.],
   [ 0.,  0.,  0.,  0.,  0.],
   [ 0.,  0.,  0.,  0.,  0.],
   [ 0.,  0.,  0.,  0.,  0.],
   [ 0.,  0.,  0.,  0.,  0.]])

I want to assign values based on index so I did this .

out = np.array([[nan,2.,4.,1.,1.],[nan,3.,4.,4.,4.]])

arr[out[0].astype(int),np.arange(len(out[0]))] = 1
arr[out[1].astype(int),np.arange(len(out[1]))] = 1

Assignment works fine if there is 0 instead of nan.

How can I skip assignment in case of nan? and Is it possible to assign values at once from a multidimensional index array rather than using for loop ?

Divakar · Accepted Answer · 2017-08-31 16:47:32Z

2

Mask it -

mask = ~np.isnan(out)
arr[out[0,mask[0]].astype(int),np.flatnonzero(mask[0])] = 1
arr[out[1,mask[1]].astype(int),np.flatnonzero(mask[1])] = 1

Sample run -

In [171]: out
Out[171]: 
array([[ nan,   2.,   4.,   1.,   1.],
       [ nan,   3.,   4.,   4.,   4.]])

In [172]: mask = ~np.isnan(out)
     ...: arr[out[0,mask[0]].astype(int),np.flatnonzero(mask[0])] = 1
     ...: arr[out[1,mask[1]].astype(int),np.flatnonzero(mask[1])] = 1
     ...: 

In [173]: arr
Out[173]: 
array([[ 0.,  0.,  0.,  0.,  0.],
       [ 0.,  0.,  0.,  1.,  1.],
       [ 0.,  1.,  0.,  0.,  0.],
       [ 0.,  1.,  0.,  0.,  0.],
       [ 0.,  0.,  1.,  1.,  1.]])

Alternative, replace the flatnonzero calls with range-masking -

r = np.arange(arr.shape[1])
arr[out[0,mask[0]].astype(int),r[mask[0]]] = 1
arr[out[1,mask[1]].astype(int),r[mask[1]]] = 1

If you are working with a lot many rows than just 2 and you want to assign them in a vectorized manner, here's one method, using linear-indexing -

n = arr.shape[1]
linear_idx = (out*n + np.arange(n))
np.put(arr, linear_idx[~np.isnan(linear_idx)].astype(int), 1)

edited Aug 31, 2017 at 16:47

answered Aug 31, 2017 at 15:22

Divakar

222k19 gold badges273 silver badges374 bronze badges

Sign up to request clarification or add additional context in comments.

3 Comments

Bharath M Shetty Over a year ago

Sir and my second question.

Divakar Over a year ago

@Bharathshetty Where's the second question?

Bharath M Shetty Over a year ago

without using arr[out[0... and arr[out[1... is it possible to write the same in one line

Collectives™ on Stack Overflow

Index into NumPy array ignoring NaNs in the indexing array

1 Answer 1

3 Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

1 Answer 1

3 Comments

Your Answer

Sign up or log in

Post as a guest

Related