Reshape an array in NumPy

Question

Consider an array of the following form (just an example):

[[ 0  1]  [ 2  3]  [ 4  5]  [ 6  7]  [ 8  9]  [10 11]  [12 13]  [14 15]  [16 17]] 

It's shape is [9,2]. Now I want to transform the array so that each column becomes a shape [3,3], like this:

[[ 0  6 12]  [ 2  8 14]  [ 4 10 16]] [[ 1  7 13]  [ 3  9 15]  [ 5 11 17]] 

The most obvious (and surely "non-pythonic") solution is to initialise an array of zeroes with the proper dimension and run two for-loops where it will be filled with data. I'm interested in a solution that is language-conform...

Answer 1

a = np.arange(18).reshape(9,2) b = a.reshape(3,3,2).swapaxes(0,2)  # a:  array([[ 0,  1],        [ 2,  3],        [ 4,  5],        [ 6,  7],        [ 8,  9],        [10, 11],        [12, 13],        [14, 15],        [16, 17]])   # b: array([[[ 0,  6, 12],         [ 2,  8, 14],         [ 4, 10, 16]],         [[ 1,  7, 13],         [ 3,  9, 15],         [ 5, 11, 17]]]) 

Answer 2

numpy has a great tool for this task ("numpy.reshape") link to reshape documentation

a = [[ 0  1]  [ 2  3]  [ 4  5]  [ 6  7]  [ 8  9]  [10 11]  [12 13]  [14 15]  [16 17]]  `numpy.reshape(a,(3,3))` 

you can also use the "-1" trick

`a = a.reshape(-1,3)` 

the "-1" is a wild card that will let the numpy algorithm decide on the number to input when the second dimension is 3

so yes.. this would also work: a = a.reshape(3,-1)

and this: a = a.reshape(-1,2) would do nothing

and this: a = a.reshape(-1,9) would change the shape to (2,9)