I have the masked array (toy example) - [1 2 3 4 -- 6]
points = [1, 2, 3, 4, 0, 6]
mask = [0, 0, 0, 0, 1, 0]
points = ma.array(points, mask=mask)
And I would like to interpolate it from 6 dimensions to any number, for example, 6. My interpolation criteria is that it will ignore masked values, and skip that index.
Undesired behaviour, for example with lin = np.linspace(0, 1, 6):
f = interp1d(lin, points, axis=0, kind='cubic')
f(lin) # [1 2 3 4 -8.8817842e-16 6]
Instead, I'm expecting it to behave like:
compressed_lin = [0, 0.2, 0.4, 0.6, 1]
compressed_points = np.array([1,2,3,4,6])
f = interp1d(compressed_lin, compressed_points, axis=0, kind='cubic')
f(lin) # [1 2 3 4 5 6]
Actual Data:
My data is in the shape of [100, 100, 2], so its not as simple as to hide the masked values from a linspace of size 100.
Note on the masked array:
I already know there is a way to do a 2d interpolation on a masked array (https://modelhelptokyo.wordpress.com/2017/10/25/how-to-interpolate-missing-values-2d-python/, Scipy interpolation with masked data?, Scipy interp2d interpolate masked fill values)
If there was a way to interpolate all of the values in the masked array to be filled, rather than a constant fill, that will solve the problem, as then I'll be dealing with a filled numpy array.
The reason I can't use compressed is it can change the order of indexes and the size of the array, e.g:
points = np.array([1, 2, 3, 4, 5, 6])
points = ma.stack([
ma.array(points, mask=[0, 0, 0, 0, 1, 0]),
ma.array(points, mask=[0, 0, 1, 0, 0, 0]),
])
print(points)
# [[1 2 3 4 -- 6]
# [1 2 -- 4 5 6]]
print(np.reshape(points.compressed(), (2, 5)))
# [[1 2 3 4 6]
# [1 2 4 5 6]]
