Numpy roll vertical in 2d array

Question

How would one (efficiently) do the following:

x = np.arange(49)
x2 = np.reshape(x, (7,7))

x2
array([[ 0,  1,  2,  3,  4,  5,  6],
       [ 7,  8,  9, 10, 11, 12, 13],
       [14, 15, 16, 17, 18, 19, 20],
       [21, 22, 23, 24, 25, 26, 27],
       [28, 29, 30, 31, 32, 33, 34],
       [35, 36, 37, 38, 39, 40, 41],
       [42, 43, 44, 45, 46, 47, 48]])

From here I want to roll a couple of things. I want to roll 0,7,14,21 etc so 14 comes to top. Then the same with 4,11,18,25 etc so 39 comes to top.
Result should be:

x2
array([[14,  1,  2,  3, 39,  5,  6],
       [21,  8,  9, 10, 46, 12, 13],
       [28, 15, 16, 17,  4, 19, 20],
       [35, 22, 23, 24, 11, 26, 27],
       [42, 29, 30, 31, 18, 33, 34],
       [ 0, 36, 37, 38, 25, 40, 41],
       [ 7, 43, 44, 45, 32, 47, 48]])

I looked up numpy.roll, here and google but couldn't find how one would do this.
For horizontal rolls, I could do:

np.roll(x2[0], 3, axis=0)

x3
array([4, 5, 6, 0, 1, 2, 3])

But how do I return the full array with this roll change as a new copy?

Nickil Maveli · Accepted Answer · 2016-11-01 13:46:14Z

6

Roll with a negative shift:

x2[:, 0] = np.roll(x2[:, 0], -2)

Roll with a positive shift:

x2[:, 4] = np.roll(x2[:, 4], 2)

gives:

>>>x2
array([[14,  1,  2,  3, 39,  5,  6],
       [21,  8,  9, 10, 46, 12, 13],
       [28, 15, 16, 17,  4, 19, 20],
       [35, 22, 23, 24, 11, 26, 27],
       [42, 29, 30, 31, 18, 33, 34],
       [ 0, 36, 37, 38, 25, 40, 41],
       [ 7, 43, 44, 45, 32, 47, 48]])

answered Nov 1, 2016 at 13:46

Nickil Maveli

29.8k10 gold badges86 silver badges88 bronze badges

Sign up to request clarification or add additional context in comments.

Comments

Divakar · Accepted Answer · 2016-11-01 14:14:37Z

Here's a way to roll multiple columns in one go with advanced-indexing -

# Params
cols = [0,4]  # Columns to be rolled
dirn = [2,-2] # Offset with direction as sign

n = x2.shape[0]
x2[:,cols] = x2[np.mod(np.arange(n)[:,None] + dirn,n),cols]

Sample run -

In [45]: x2
Out[45]: 
array([[ 0,  1,  2,  3,  4,  5,  6],
       [ 7,  8,  9, 10, 11, 12, 13],
       [14, 15, 16, 17, 18, 19, 20],
       [21, 22, 23, 24, 25, 26, 27],
       [28, 29, 30, 31, 32, 33, 34],
       [35, 36, 37, 38, 39, 40, 41],
       [42, 43, 44, 45, 46, 47, 48]])

In [46]: cols = [0,4,5]  # Columns to be rolled
    ...: dirn = [2,-2,4] # Offset with direction as sign
    ...: n = x2.shape[0]
    ...: x2[:,cols] = x2[np.mod(np.arange(n)[:,None] + dirn,n),cols]
    ...: 

In [47]: x2  # Three columns rolled
Out[47]: 
array([[14,  1,  2,  3, 39, 33,  6],
       [21,  8,  9, 10, 46, 40, 13],
       [28, 15, 16, 17,  4, 47, 20],
       [35, 22, 23, 24, 11,  5, 27],
       [42, 29, 30, 31, 18, 12, 34],
       [ 0, 36, 37, 38, 25, 19, 41],
       [ 7, 43, 44, 45, 32, 26, 48]])

dnalow · Accepted Answer · 2016-11-01 12:26:24Z

0

You have to overwrite the column

e.g.:

x2[:,0] = np.roll(x2[:,0], 3)

answered Nov 1, 2016 at 12:26

dnalow

9844 silver badges14 bronze badges

Comments

leenremm · Accepted Answer · 2020-04-05 16:21:07Z

0

See here a useful method for shifting a 2D array in all 4 directions (up, down, left, right):

def image_shift_roll(img, x_shift, y_roll):
    img_roll = img.copy()
    img_roll = np.roll(img_roll, -y_roll, axis = 0)    # Positive y rolls up
    img_roll = np.roll(img_roll, x_roll, axis = 1)     # Positive x rolls right
    return img_roll

answered Apr 5, 2020 at 16:21

leenremm

1,47418 silver badges22 bronze badges

Collectives™ on Stack Overflow

Numpy roll vertical in 2d array

4 Answers 4

Comments

Comments

Comments

Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

4 Answers 4

Comments

Comments

Comments

Comments

Your Answer

Sign up or log in

Post as a guest

Related