Repository navigation
Warn re: Axes3D constructor behavior change in mpl3.4 #18939
Description
According to git bisect, 261f706 by @anntzer broke the following code, adapted from the examples -- it produces a blank / empty plot:
from mpl_toolkits.mplot3d import Axes3D
import matplotlib.pyplot as plt
import numpy as np
fig = plt.figure()
ax = Axes3D(fig)
# ax = fig.gca(projection='3d')
X = np.arange(-5, 5, 0.25)
Y = np.arange(-5, 5, 0.25)
X, Y = np.meshgrid(X, Y)
R = np.sqrt(X**2 + Y**2)
Z = np.sin(R)
surf = ax.plot_surface(X, Y, Z)
If you comment out the ax = Axes3D(fig) line and use ax = fig.gca(projection='3d'), it works fine. It also works fine on the preceding commit 34f99a9.
Activity
anntzer commentedon Nov 12, 2020
That was actually mostly discussed in #18356. The change was definitely intentional (as both #18356 and #18564 needed it) and documented at https://github.com/matplotlib/matplotlib/blob/master/doc/api/next_api_changes/behavior/18564-AL.rst; do you want to discuss reverting it, or emit a warning, or something else?
larsoner commentedon Nov 12, 2020
Ahh I just didn't notice that change. A warning that it is no longer added would have made finding the problem immediately apparent, so +1 for that solution if it's tractable.
anntzer commentedon Nov 12, 2020
I think we can (temporarily) add a warning in
Axes3D.__del__checking whether the axes has ever been added to a figure.[-]BUG: Regression in Axes3D use[/-][+]Warn re: Axes3D constructor behavior change in mpl3.4[/+]larsoner commentedon Nov 12, 2020
Actually a simpler approach would be just to add a
add=Trueargument toAxes3D. Then people who used it like I did get their behavior preserved, and internallymatplotlibcan doadd=False. Then there isn't any need for the release notes or user changes. And hopefully it's just two linesif add: fig.add_subplot(self)or so.MNT: Restore and deprecate auto-adding Axes3D to figure on init
MNT: Restore and deprecate auto-adding Axes3D to figure on init
MNT: Restore and deprecate auto-adding Axes3D to figure on init