Repository navigation
[Bug]: pcolormesh with log norm on data with zeros is broken, unless colorbar is added #32599
Description
Bug summary
I'm plotting 2D array with pcolormesh and:
- use logarithmic normalization (
norm="log") array contains zeros (unity matrix in the MRE example)- array containing only one non-zero value (e.g. unity)
Code for reproduction
import numpy as np
import matplotlib.pyplot as plt
fig, axes = plt.subplots(figsize=(8, 4), ncols=2)
for i, ax in enumerate(axes):
mesh = ax.pcolormesh(np.eye(3), norm="log")
if i == 1:
fig.colorbar(mesh, ax=ax)
fig.savefig("mre.png")Actual outcome
The resulting plot has all bins colored the same color (cmap(0)) (left). However, if I add colorbar to the axes (right), the result is correct.
Expected outcome
Correct result regarding of colorbar presence (lognorm color applied to non-zero bins, zero ones are white/blank).
Additional information
No response
Operating system
OS/X
Matplotlib Version
3.11.2, also tested in 3.10.5
Matplotlib Backend
macosx
Python version
Python 3.14.9
Jupyter version
No response
Installation
pip
Activity
nj-vs-vh commentedon Nov 14, 2026
I didn't find this issue reported but it seems plausibly related to #8540
nj-vs-vh commentedon Nov 14, 2026
UPD: when the array has >1 non-zero unique value, everything is consistent
rcomer commentedon Nov 14, 2026
Thanks for the report. It seems that colorbar somehow modifies the vmin and vmax of the norm, which I assume is deliberate. We have
Without colorbar:
With colorbar:
I think the output masks should match for these two cases, even if the norms' vmin and vmax remain different.