Modify 3D numpy array using interpolation

Question

I have a 3D array that I want to interpolate the np.nan values along the z dimension, and I just want the changes to modify my existing array. However, the changes seems not to be working. I have a test array with dimension (3,3,3) with nan values. I am accessing the z dimension and perform interpolation. I would like to have is that nan values will be replaced by interpolated values along the z dimension.

def arr_interp(array):
    arrN=np.array(array)
    ix=-np.isnan(arrN)
    idxTrue=ix.nonzero()[0]
    valueTrue=arrN[-np.isnan(arrN)]
    countFalse=np.isnan(arrN).ravel().nonzero()[0]
    arrN[np.isnan(arrN)]=np.interp(countFalse,idxTrue,valueTrue)
    # I got this function from another forum


In [177]:arrStack
Out[178]: 
array([[[ 0.59016759,         nan,  0.88641936],
        [ 0.94884139,         nan,  0.89443098],
        [ 0.7730535 ,         nan,  0.86262287]],

       [[ 0.106567  ,         nan,  0.83757824],
        [ 0.21794018,         nan,  0.66128902],
        [ 0.21961069,         nan,  0.28654693]],

       [[ 0.26142901,         nan,  0.95653887],
        [ 0.19616204,  0.45      ,  0.61531475],
        [ 0.78262652,  0.76      ,  0.47786104]]])

In [181]: arrStack[0][0]
Out[182]: array([ 0.59016759,         nan,  0.88641936])

In [183]: arr_interp(arrStack[0][0])

In [184]: arrStack
Out[184]: 
array([[[ 0.59016759,         nan,  0.88641936],
        [ 0.94884139,         nan,  0.89443098],
        [ 0.7730535 ,         nan,  0.86262287]],

       [[ 0.106567  ,         nan,  0.83757824],
        [ 0.21794018,         nan,  0.66128902],
        [ 0.21961069,         nan,  0.28654693]],

       [[ 0.26142901,         nan,  0.95653887],
        [ 0.19616204,  0.45      ,  0.61531475],
        [ 0.78262652,  0.76      ,  0.47786104]]])

Answer 1

The problem is in your second line of code

arrN=np.array(array)

What this does is create a copy of your input array since the standard behaviour is np.array(x, copy=True). This way you are interpolating the copy instead of the original array.

If you want to modify the existing array in place just change it to:

arrN=np.array(array, copy=False)

This way arrN points to the original input array.