基于布尔数组将列插入到numpy数组中

Question

Apologies if this question has an elementary answer, but I have not been able to find one yet.

I have some data that was originally a numpy array of shape (N, M) , but where certain columns have then been removed. For illustration lets say M=6 and 2 columns where removed, leaving a (N, 4) array. What I also have is an array denoting whether a column was kept or not with a Boolean value ( True if kept, False if not), eg array([False, True, True, False, True, True]) .

What I would like to do would be to reconstruct an (N, 6) array from the (N, 4) array and the boolean markers, with the columns reintroduced at the right index (filled with zeros).

Any help on the requisite slicing etc approaches would be valued!

Answer 1

Let's try this:

# original array -- for references only
arr = np.arange(12).reshape(-1,6)

# keep indexes
keep = np.array([False,  True,  True,  False,  True,  True])

# array after removing the columns
removed_arr = arr[:,keep]

# output array
out = np.zeros((removed_arr.shape[0], len(keep)), dtype=removed_arr.dtype)

# copy the values of `removed_arr` to output
out[:, keep] = removed_arr

Output:

array([[ 0,  1,  2,  0,  4,  5],
       [ 0,  7,  8,  0, 10, 11]])