Correctly scale quadmesh linewidth to dpi #3793
Conversation
13901f9 to
9dc7098
Compare
9dc7098 to
27186be
Compare
| fig.add_subplot(111) | ||
| np.random.seed(0) | ||
| r = np.random.random((4, 4)) | ||
| plt.contourf(r, interpolation="nearest", extend="both") |
There was a problem hiding this comment.
Contourf doesn't have an "interpolation" kwarg.
There was a problem hiding this comment.
It always bugs me that you can pass any kwarg to contour/contourf - we should fix that someday.
|
Maybe the test could be using the quadmesh more directly, with something like plt.pcolormesh(np.array([[1, 2], [3, 4]]),
cmap=plt.get_cmap('gray'),
edgecolors='r',
lw=10)I think that would do it. |
|
The test failure is in matplotlib.tests.test_axes.test_pcolormesh.test; regenerating that test image should take care of it. |
| @image_comparison(baseline_images=['quadmesh_linewidth'], | ||
| remove_text=True, | ||
| extensions=['png'], | ||
| savefig_kwarg={'dpi': 400}) |
There was a problem hiding this comment.
Do we need an image quite this big? Can 8000 we work with something smaller?
There was a problem hiding this comment.
This bug is only visible at higher dpis, which is why I did it this way. But given the other test that failed here (test_axes:pcolormesh), it's probably best to just update the baseline for that test and not add a new test at all.
27186be to
f2d4631
Compare
|
Thanks. On further thought, I think it's best to just update the pcolormesh test baseline image and let that be the catch for this bug (rather than added a new test and increasing the size of the repo). |
Correctly scale quadmesh linewidth to dpi
As this touches C++ code, it will have a non-straightforward merge to master...