python - sort 3D array by the sum of the second value -

i'm trying sort 3d array , having little success. have read various methods on how done 2 dimensions, can't seem scale up.

i have array like:

[[[0, 25], [1, 20], [2, 80], [0, 0], [0, 0], [0, 0]],  [[0, 80], [0, 20], [1, 25], [1, 40], [2, 99], [0, 0]],  [[0, 99], [1, 40], [2, 80], [0, 0], [0, 0], [0, 0]]] 

i want sorted sum of second value. criteria use order sum of second value. e.g:

 25+20+80 = 125  80+20+25+40+99 = 264  99+40+80 = 219 

so:

 [[[0, 80], [0, 20], [1, 25], [1, 40], [2, 99], [0, 0]],   [[0, 99], [1, 40], [2, 80], [0, 0], [0, 0], [0, 0]]],   [[0, 25], [1, 20], [2, 80], [0, 0], [0, 0], [0, 0]]] 

could me ?

i think understand need. suppose array stored in a.

>>> out[19]:  array([[[ 0, 25],         [ 1, 20],         [ 2, 80],         [ 0,  0],         [ 0,  0],         [ 0,  0]],         [[ 0, 80],         [ 0, 20],         [ 1, 25],         [ 1, 40],         [ 2, 99],         [ 0,  0]],         [[ 0, 99],         [ 1, 40],         [ 2, 80],         [ 0,  0],         [ 0,  0],         [ 0,  0]]]) 

a[:, :, 1] accesses second element of innermost dimension, can sum on (on horizontal axis, i.e. axis=1) values need sorting:

>>> a[:, :, 1] out[20]:  array([[25, 20, 80,  0,  0,  0],        [80, 20, 25, 40, 99,  0],        [99, 40, 80,  0,  0,  0]]) >>> b = np.sum(a[:, :, 1], axis=1) >>> b out[21]: array([125, 264, 219]) 

use np.argsort sorted indices, rather sorted array:

>>> = np.argsort(b)[::-1] # reverse array descending order >>> out[23]: array([1, 2, 0]) 

finally, access array using indices, desired:

>>> a[i, :, :] out[25]:  array([[[ 0, 80],         [ 0, 20],         [ 1, 25],         [ 1, 40],         [ 2, 99],         [ 0,  0]],         [[ 0, 99],         [ 1, 40],         [ 2, 80],         [ 0,  0],         [ 0,  0],         [ 0,  0]],         [[ 0, 25],         [ 1, 20],         [ 2, 80],         [ 0,  0],         [ 0,  0],         [ 0,  0]]]) 

hope helps!