ImageDraw.rectangle off by a factor of 2 for uint16 type images

[TAdeJong]

What did you do?

Draw a rectangle of size 6*6 in a image with dtype np.uint16, per following code:

from PIL import ImageDraw, Image
import matplotlib.pyplot as plt
import numpy as np

I = Image.fromarray(np.zeros((8,8), dtype=np.uint16))
d = ImageDraw.Draw(I)
d.rectangle([0,0,5,5], fill='white')
np.asarray(I)

What did you expect to happen?

An output similar to the case of a float image or a uint8 image, (taking into account #367 and #1668), with a square white triangle.

What actually happened?

The output is as follows:

array([[65535, 65535, 65535,     0,     0,     0,     0,     0],
       [65535, 65535, 65535,     0,     0,     0,     0,     0],
       [65535, 65535, 65535,     0,     0,     0,     0,     0],
       [65535, 65535, 65535,     0,     0,     0,     0,     0],
       [65535, 65535, 65535,     0,     0,     0,     0,     0],
       [65535, 65535, 65535,     0,     0,     0,     0,     0],
       [    0,     0,     0,     0,     0,     0,     0,     0],
       [    0,     0,     0,     0,     0,     0,     0,     0]],
      dtype=uint16)

So the rectangle is only half the width compared to the expected size. This only happens for uint16 images as far as I checked.

What are your OS, Python and Pillow versions?

• OS: Windows 10
• Python: Anaconda 3.6.8
• Pillow: 5.4.1