[Bug] Incorrect ytick labels in shared subplot with log yscale

Development
1

Matplotlib is weak in handling automatic ticks and tick labels with log scale axes. I refuse to use the complicated

mticker

and resort instead to the much simpler

yticks = …
yticklabels = [str(yt) for yt in yticks]
plt.yticks(yticks)
plt.gca().set_yticklabels(yticklabels)

This works OK (especially in a function I created), except when I use subplots with shared axes. Here is an example.

from matplotlib import pyplot as plt
import numpy as np
x = np.arange(0,10)
y0 = x
y1 = x * (1+0.2*np.random.randn(10))
fig, ax = plt.subplots(1,2,sharey=True,squeeze=True,figsize=[8,5])
plt.subplots_adjust(wspace=0)
ax[0].scatter(x,y0,c=‘k’)
ax[0].set_xlabel(‘x’)
ax[0].set_ylabel(‘x’)
ax[0].set_yscale(‘log’)
yticks = [1,2,5,10]
yticklabels = [str(yt) for yt in yticks]
ax[0].set_yticks(yticks)
ax[0].set_yticklabels(yticklabels)
ax[0].set_title(‘y-axis labels should be at 1, 2, 5 and 10’)
ax[1].scatter(x,y1,c=‘k’)
ax[1].set_xlabel(‘x again’)
ax[1].set_ylabel(‘x with some randomness’)
ax[1].yaxis.set_label_position(‘right’)
ax[1].set_yscale(‘log’)
ax[1].set_yticks(yticks)

The output shows large taxis ticks at the desired positions, but the y-axis labels are only displayed at 0.1, 1, 10, whereas they were requested at closer spacings. This seems to me to be a bug. It goes away if

sharey=True

is removed. In this example, I removed my matplotlibrc file before making the figure.

Is this bug easy to fix?

Python 3.9.16
Matplotlib 3.7.0
MacOS 13.6.1

2

Remove these lines:

ax[1].set_yscale(‘log’)
ax[1].set_yticks(yticks)

Since the y-axis is shared, when you call set_yscale on the second axes, it overrides what you already set on the first axes.

3

Thanks Ruth! This works indeed. I should have realized that sharey=True would not need new yticks and yscale.

1 Like