Skip to content

Scale an image to the axes edges regardless of whether it is being plotted with a nonaffine transform or an affine transform#30175

Closed
ayshih wants to merge 2 commits intomatplotlib:mainfrom
ayshih:nonaffine_scaling
Closed

Scale an image to the axes edges regardless of whether it is being plotted with a nonaffine transform or an affine transform#30175
ayshih wants to merge 2 commits intomatplotlib:mainfrom
ayshih:nonaffine_scaling

Conversation

@ayshih
Copy link
Contributor

@ayshih ayshih commented Jun 14, 2025

PR summary

This PR fixes a difference between plotting an image with a nonaffine transform versus with an affine transform. When plotting with an affine transform, the transform for rendering gets an additional transform step to scale the image to the axes edges in the situation where the axes bbox is not a whole number of pixels. However, when plotting with a nonaffine transform, this additional scaling step is skipped, and instead the fractional pixel at the end is clipped off. I can't think of any reason why this difference is desirable, so this PR removes the different handling of a nonaffine transform versus an affine transform.

Before this PR

Here's an example image before this PR and generating script. Note that the image plotted using the nonaffine transform has a obvious gap to the axis on the right edge due to the fractional pixel. The discrepancy on the top edge is not quite as obvious.
Figure_1

import numpy as np

import matplotlib.pyplot as plt
from matplotlib.transforms import Transform

# Create a nonaffine identity transform
class NonAffineIdentityTransform(Transform):
    input_dims = 2
    output_dims = 2

    def inverted(self):
       return self
nonaffine_transform = NonAffineIdentityTransform()

data = np.arange(16).reshape((4, 4)) % 3

fig, axs = plt.subplots(2, figsize=(6, 3), dpi=100)

# Choose axes bboxs with with fractional pixels
axs[0].set_position([0.1, 0.15, 200.75 / fig.bbox.width, 200.75 / fig.bbox.height])
axs[1].set_position([0.6, 0.15, 200.75 / fig.bbox.width, 200.75 / fig.bbox.height])

im0 = axs[0].imshow(data)
im1 = axs[1].imshow(data, transform=nonaffine_transform + axs[1].transData)

axs[0].spines[:].set_ls((0, (5, 5)))
axs[1].spines[:].set_ls((0, (5, 5)))

axs[0].set_title("Affine")
axs[1].set_title("Nonaffine")

plt.show()

After this PR

after

PR checklist

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants