load_image data has incorrect shape ?

[euri10]

Describe the bug Loading an image it seems to me the _data numpy array has the wrong shape, at least it feels this way to me but I can be off.

To Reproduce

from p5 import size, load_image, load_pixels, image, run
from p5.core.image import PImage

akira_w = 200
akira_h = 89


def setup():
    global img, sorted
    size(2 * akira_w, akira_h)
    img = load_image("akira.png")
    sorted = PImage(akira_w, akira_h)
    sorted.load_pixels()
    with load_pixels():
        # fillme = np.random.random_integers(0, 255, (240,240,4))
        fillme = img._data
        print(fillme.shape)
        for x in range(akira_w):
            for y in range(akira_h):
                sorted[x, y] = fillme[y][x]


def draw():
    image(img, (0, 0), (akira_w, akira_h))
    image(sorted, (akira_w, 0), (akira_w, akira_h))


if __name__ == "__main__":
    run()

this code ends up with this image:

it feels akward to have to reverse x and y above in sorted[x, y] = fillme[y][x]

Expected behavior

I would expect to write sorted[x, y] = fillme[x][y] to have the image copied properly

Screenshots see above

System information:

• p5 release (version number or latest commit): 0.7.0
• Python version: 3.8.5
• Operating system: debian buster

Additional context Add any other context about the problem here.

[jeremydouglass]

I see what you mean: [y][x] vs [x][y].

Is _data an implementation detail to be rationalized -- or is it part of the API?

At a quick look, this might have something to do with PIL Image? At any rate, some internal reshape mapping seems to be intentional:

https://github.com/p5py/p5/blob/d7a160ac66361e06b6db70a0436cb365693ebaf3/p5/core/image.py#L179

[petiesmo]

Numpy doesn’t have a sense of ‘direction’, just axis 1, axis 2, .... axis n.

Numpy expects position as [row][col]. On the screen/canvas, a pixels is [y rows down][x cola across]. Hence, [y][x]

On Sat, Sep 12, 2020 at 6:08 PM Jeremy Douglass [email protected] wrote:

I see what you mean: [y][x] vs [x][y].

Is _data an implementation detail to be rationalized -- or is it part of the API?

At a quick look, this might have something to do with PIL Image? At any rate, some internal reshape mapping seems to be intentional:

https://github.com/p5py/p5/blob/d7a160ac66361e06b6db70a0436cb365693ebaf3/p5/core/image.py#L179

— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/p5py/p5/issues/243#issuecomment-691577203, or unsubscribe https://github.com/notifications/unsubscribe-auth/AP6TTJ4ZG7FENRV6NU3X5JLSFQEO7ANCNFSM4QMAUMAQ .