ax.set_zorder not working

[kinyatoride]

Description

ax.set_zorder does not work in proplot.
This becomes an issue when we want to change the zorders of twinx axes.

Steps to reproduce

import numpy as np
import pandas as pd
import proplot as pplt

x = np.linspace(0, 2*np.pi, 100)

fig, ax1 = pplt.subplots()
ax1.plot(np.sin(x), lw=3, c='b')

ax2 = ax1.twinx()
ax2.plot(np.sin(2*x), lw=3, c='r')

ax1.set_zorder(ax2.get_zorder()+1)
ax1.patch.set_visible(False)

Expected behavior: [What you expected to happen]
Red line to be placed beneath the blue line

Actual behavior: [What actually happened]

Equivalent steps in matplotlib

import numpy as np
import pandas as pd
import matplotlib.pyplot as plt

x = np.linspace(0, 2*np.pi, 100)

fig, ax1 = plt.subplots()
ax1.plot(np.sin(x), lw=3, c='b')

ax2 = ax1.twinx()
ax2.plot(np.sin(2*x), lw=3, c='r')

ax1.set_zorder(ax2.get_zorder()+1)
ax1.patch.set_visible(False)

Proplot version

3.4.3
0.9.4

[lukelbd]

It seems this is because (unlike matplotlib) proplot implements twin axes as "children" of the original axes rather than separate figure axes. This approach simplifies the tight layout algorithm, but means that the drawing order of the twin axes is controlled by the zorder of content inside the original axes rather than the original axes itself. So in principle, you should just need to set zorder below that of the blue line with e.g. ax1.twinx(zorder=1) (see below). I should have explained this in the documentation -- will add a disclaimer.

Your example also directed me to a bug -- zorder=1 only works in matplotlib < 3.4. In matplotlib >= 3.4, it seems the twin axes is getting drawn twice, once as a "child" of the original axes and once as a separate figure axes. So, for your purposes you need to use zorder=-1 instead. I'll fix this in the next version.

import numpy as np
import proplot as pplt

x = np.linspace(0, 2*np.pi, 100)

fig, ax1 = pplt.subplots()  # axes with default zorder 0
ax1.plot(np.sin(x), lw=3, c='b')  # line with default zorder 2
ax2 = ax1.twinx(zorder=1)  # less than the line, more than the subplot
ax2.plot(np.sin(2*x), lw=3, c='r')

[lukelbd]

This is now fixed -- you should get the above result with zorder=1 in matplotlib 3.4+, and I've added a disclaimer to the documentation noting the different zorder behavior here.