How to make a checkerboard in numpy?

Question

I'm using numpy to initialize a pixel array to a gray checkerboard (the classic representation for "no pixels", or transparent). It seems like there ought to be a whizzy way to do it with numpy's amazing array assignment/slicing/dicing operations, but this is the best I've come up with:

w, h = 600, 800
sq = 15    # width of each checker-square
self.pix = numpy.zeros((w, h, 3), dtype=numpy.uint8)
# Make a checkerboard
row = [[(0x99,0x99,0x99),(0xAA,0xAA,0xAA)][(i//sq)%2] for i in range(w)]
self.pix[[i for i in range(h) if (i//sq)%2 == 0]] = row
row = [[(0xAA,0xAA,0xAA),(0x99,0x99,0x99)][(i//sq)%2] for i in range(w)]
self.pix[[i for i in range(h) if (i//sq)%2 == 1]] = row

It works, but I was hoping for something simpler.

Answer 1

Can't you use hstack and vstack? See here. Like this:

>>> import numpy as np
>>> b = np.array([0]*4)
>>> b.shape = (2,2)
>>> w = b + 0xAA
>>> r1 = np.hstack((b,w,b,w,b,w,b))
>>> r2 = np.hstack((w,b,w,b,w,b,w))
>>> board = np.vstack((r1,r2,r1,r2,r1,r2,r1))

Answer 2

import numpy as NP

def checkerboard(w, h) :
    re = NP.r_[ w*[0,1] ]
    ro = NP.r_[ w*[1,0] ]
    return NP.row_stack(h*(re, ro))


AB = checkerboard(5, 10)

Answer 3

I'm not sure if this is better than what I had:

c = numpy.fromfunction(lambda x,y: ((x//sq) + (y//sq)) % 2, (w,h))
self.chex = numpy.array((w,h,3))
self.chex[c == 0] = (0xAA, 0xAA, 0xAA)
self.chex[c == 1] = (0x99, 0x99, 0x99)

Answer 4

Ned, I like your answer; it's quite readable. However, here's another way to do it using ogrid which is a bit faster:

import numpy as np
import Image

w,h = 600,800
sq=15
color1=(0xFF,0x80,0x00)
color2=(0x80,0xFF,0x00)
def use_ogrid():
    coords=np.ogrid[0:w,0:h]
    idx=(coords[0]//sq+coords[1]//sq)%2
    vals=np.array([color1,color2],dtype=np.uint8)
    img=vals[idx]
    return img

def use_fromfunction():
    img = np.zeros((w,h,3), dtype=np.uint8)
    c = np.fromfunction(lambda x,y: ((x//sq) + (y//sq)) % 2, (w,h))
    img[c==0]=color1
    img[c==1]=color2
    return img

if __name__=='__main__':
    for f in (use_ogrid,use_fromfunction):
        img=f()
        pilImage = Image.fromarray(img, 'RGB')
        pilImage.save('{0}.png'.format(f.func_name))

Here are the timeit results:

% python -mtimeit -s"import test" "test.use_fromfunction()"
10 loops, best of 3: 307 msec per loop
% python -mtimeit -s"import test" "test.use_ogrid()"
10 loops, best of 3: 129 msec per loop