Bug summary

Run the code below, which uses set.xscale("log") and NonUniformImage. Choose the move tool (Arrows in all 4 directions) and try to pan the image around. Instead of panning, image distorts.

Code for reproduction

import numpy as np
import matplotlib.pyplot as plt
from matplotlib.image import NonUniformImage

chg=np.logspace(0,1,10)
chg1=chg[:,None]
chg2=chg[None,:]
g=chg1+chg2
g+=50*(np.add.outer(np.arange(10),np.arange(10))%2)
#checkerboard, to make effect clearer
fig,ax=plt.subplots()
ax.set_xscale("log")
ax.set_yscale("log")
im=NonUniformImage(ax,extent=[chg[0],chg[-1],chg[0],chg[-1]],origin="lower")


im.set_data(np.linspace(1,10,10),np.linspace(1,10,10),g)
ax.add_image(im)

plt.show()

Actual outcome

These are supposed to be the same image. I just used the pan tool.
But the aspect ratio has changed. And the boundary of the rightmost colour change has gone from 0.75 to 0.5.

Clicking and dragging produces behaviour that is visually obviously not-right.

Expected outcome

The image should pan around.

Additional information

No bug appears when I use a plt.imshow on the log scale axis. Or when I use the same NonUniformImage code without the logscale axis. The bug only appears with all 3 conditions are met.

I was trying to find a sensible way to display an image on log-log axes with constant screen space per pixel, when I discovered this bug.

Operating system

Ubuntu

Matplotlib Version

3.8.0

Matplotlib Backend

QtAgg

Python version

Python 3.11.5

Jupyter version

No response

Installation

conda