CODEWITHSUNDEEP
pythonnumpy
Emily
Emily

Reputation: 865

Save.npy masked array to a .npy array with NaNs where mask == True

Is there a way to convert a masked 3D numpy array to a numpy array with NaNs in place of the mask? This way I can easily write the numpy array out using np.save. The alternative is to find a way to write out the masked array with some clear indicator for elements that are masked. I have tried:

a = np.ma.zeros((500, 500))
a.dump('test')

but I need the file to be in a format so it can be read into R. Thanks.

Upvotes: 4

Views: 3165

Answers (2)

unutbu
unutbu

Reputation: 879701

A scan of the masked array operations page shows np.ma.filled does what you are looking for. For example,

import numpy as np

arr = np.arange(2*3*4).reshape(2,3,4).astype(float)
mask = arr % 5 == 0

marr = np.ma.array(arr, mask=mask)
print(np.ma.filled(marr, np.nan))

yields

[[[ nan   1.   2.   3.]
  [  4.  nan   6.   7.]
  [  8.   9.  nan  11.]]

 [[ 12.  13.  14.  nan]
  [ 16.  17.  18.  19.]
  [ nan  21.  22.  23.]]]

Alternatively, you could use the masked array's filled method. marr.filled(np.nan) is equivalent to np.ma.filled(marr, np.nan).

Upvotes: 7

James O Connor
James O Connor

Reputation: 81

You can multiply the mask by the data so that all masked elements will equal zero, eg:

a=np.random.rand(500,500)
b=ma.masked_greater(a, .5)
result=b.data*b.mask

This can then be converted to nan using 

result[result==0] = np.nan

Upvotes: 1

Related Questions