I have the following code:

zero_indices = np.where(A==0) 
nonzero_indices = np.where(A!=0)

But I want to find zero_indices and nonzero_indices with just a single call of np.where.

One way I found to do this is:

zero_indices = np.where(array == 0)
non_zero_indices = np.setdiff1d(np.arange(array.size), zero_indices)

However, when testing on large arrays, I found that this is actually 10x slower than just calling np.where twice.

Is there a more efficient way to do this?