Is it possible to reshape a np.array() and, in case of inconsistency of the new shape, fill the empty spaces with NaN?
Ex:
arr = np.array([1,2,3,4,5,6])
Target, for instance a 2x4 Matrix:
[1 2 3 4]
[5 6 NaN NaN]
I need this to bypass the error: ValueError: cannot reshape array of size 6 into shape (2,4)
We'll use np.pad first, then reshape:
m, n = 2, 4
np.pad(arr.astype(float), (0, m*n - arr.size),
mode='constant', constant_values=np.nan).reshape(m,n)
array([[ 1., 2., 3., 4.],
[ 5., 6., nan, nan]])
The assumption here is that arr is a 1D array. Add an assertion before this code to fail on unexpected cases.
Lots of ways of doing this, but (nearly) all amount to creating a new array of the desired shape, and filling values:
In [50]: arr = np.array([1,2,3,4,5,6])
In [51]: res = np.full((2,4), np.nan)
In [52]: res
Out[52]:
array([[nan, nan, nan, nan],
[nan, nan, nan, nan]])
In [53]: res.flat[:len(arr)]=arr
In [54]: res
Out[54]:
array([[ 1., 2., 3., 4.],
[ 5., 6., nan, nan]])
I used flat to treat res as a 1d array for copy purposes.
An exception is the resize method, but that fills with 0s. And doesn't change the dtype to allow for float nan:
In [55]: arr.resize(2,4)
In [56]: arr
Out[56]:
array([[1, 2, 3, 4],
[5, 6, 0, 0]])
One possible solution :
convert array to float (nan is a float type)
arr = np.array([1,2,3,4,5,6]).astype(float)
resize data to new shape
arr = np.resize(arr, (2,4))
print(arr)
array([[1., 2., 3., 4.],
[5., 6., 1., 2.]])
replace last two entries with np.NaN
arr[-1,-2:] = np.NaN
print(arr)
array([[ 1., 2., 3., 4.],
[ 5., 6., nan, nan]])