Description
Bug summary
Shared axes normally share the same scale, but it is inconsistent whether clear()ing a shared axes resets the scale on that axes, or on the shared axes too.
Code for reproduction
# Create 6 pairs of shared axes,
# set either of them to xscale("log"),
# then do noting or clear() either of them.
from pylab import *
sfs[0].suptitle("set primary")
ax0 = sfs[0].add_subplot(211)
ax1 = sfs[0].add_subplot(212, sharex=ax0)
ax0.set_xscale("log")
sfs[1].suptitle("set secondary")
ax0 = sfs[1].add_subplot(211)
ax1 = sfs[1].add_subplot(212, sharex=ax0)
ax1.set_xscale("log")
sfs[2].suptitle("set primary\nclear secondary")
ax0 = sfs[2].add_subplot(211)
ax1 = sfs[2].add_subplot(212, sharex=ax0)
ax0.set_xscale("log"); ax0.clear()
sfs[3].suptitle("set primary\nclear secondary")
ax0 = sfs[3].add_subplot(211)
ax1 = sfs[3].add_subplot(212, sharex=ax0)
ax0.set_xscale("log"); ax1.clear()
sfs[4].suptitle("set secondary\nclear primary")
ax0 = sfs[4].add_subplot(211)
ax1 = sfs[4].add_subplot(212, sharex=ax0)
ax1.set_xscale("log"); ax0.clear()
sfs[5].suptitle("set secondary\nclear secondary")
ax0 = sfs[5].add_subplot(211)
ax1 = sfs[5].add_subplot(212, sharex=ax0)
ax1.set_xscale("log"); ax1.clear()
for sf in sfs:
for ax in sf.axes:
ax.text(.5, .5, ax.get_xscale(), transform=ax.transAxes,
ha="center", va="center")
ax.set_yticks([])
show()Actual outcome
Setting the scale on either axes of a pair also sets the scale on the other axes.
Clearing the primary axes changes it, but doesn't affect the secondary axes.
Clearing the secondary axes doesn't affect either the primary or the secondary axes.
Expected outcome
Not sure what the best behavior is, but ending with a pair of shared axes with different scales is probably not good (e.g., because they also share locators and tickers, but linear locators don't play well with log scales as the former don't know anything about avoiding negative values -- this is more or less the root cause of #9970.).
Additional information
No response
Operating system
Arch
Matplotlib Version
3.6.0.dev2729+g446de7bbb6
Matplotlib Backend
mplcairo
Python version
Python 3.10.5
Jupyter version
ENOSUCHLIB
Installation
git checkout