Path3DCollection from 3D scatter cannot set_color #13035

Bug report

Bug summary
The set_color method on the Path3DCollection object, such as the one returned by 3D scatter plot, does not set the color. I am filing this as a new issue because

• all previous instances reported the issue as related to the figure not being updated / redrawn
• previous reports date back more than 3 years and to version 1.3.1
• the MWE here shows the issue is present even before the first display (show call) of the figure
• this report describes some other anomalies related to alpha that might favorable to handle together

Code for reproduction

# Can we change color after plotting? NO, independently of the color format.
# Note: in 2D set_color() works as expected.
import matplotlib.pyplot as plt
from mpl_toolkits.mplot3d import Axes3D
fig = plt.figure()
ax = fig.gca(projection = "3d")
sc1 = ax.scatter([0,1], [0,1], [1,1], color = (1,0,0,1), depthshade = False)
sc2 = ax.scatter([0,1], [0,1], [2,2], color = (1,0,0,1), alpha=0.3, depthshade = True)
sc3 = ax.scatter([0,1], [0,1], [3,3], color = (1,0,0,1), alpha=1.0, depthshade = False)
sc4 = ax.scatter([0,1], [0,1], [4,4], color = (1,0,0,0.3), alpha=1.0, depthshade = True)
sc1.set_color('blue')
sc2.set_color(['blue', 'blue'])
sc3.set_color([(0,0,1), (0,0,1)])
sc4.set_color([(0,0,1,1), (0,0,1,1)])
sc1.changed()  # workaround https://stackoverflow.com/a/8974087 does not work here 
ax.set_zticks([1,2,3,4])
ax.set_zticklabels(['"blue"', '[b,b]', 'RGB', 'RGBA'])
fig.show()

Actual outcome

While all the scatter plots are expected to have blue color.

Related questions

If the issue is coming from how Axes3D.scatter produces the Path3DCollection, it might be worthy to note the following on the color-related alpha channel.

Context: Does depthshade work with alpha value?

• NO, depthshade is negatively overridden if alpha is provided, in addition, channel A from RGBA colors is omitted.
• Remark: Since I did not find any note on this alpha vs channel A interaction in the docs, my expectation was to display using the product of alpha and A. Remark holds for 2D as well.

fig = plt.figure()
ax = fig.gca(projection = "3d")
sc1 = ax.scatter([0,1], [0,1], [1,1], color = (1,0,0,1), depthshade = False)
sc2 = ax.scatter([0,1], [0,1], [2,2], color = (1,0,0,1), alpha=0.3, depthshade = True)
sc3 = ax.scatter([0,1], [0,1], [3,3], color = (1,0,0,1), alpha=1.0, depthshade = False)
sc4 = ax.scatter([0,1], [0,1], [4,4], color = (1,0,0,0.3), alpha=1.0, depthshade = True)
ax.set_zticks([1,2,3,4])
ax.set_zticklabels(['A=N a=1', 'A=1 a=.3', 'A=1 a=1', 'A=.3 a=1'])
fig.show()

Question: Can we change alpha after plotting?

• YES, independently of how alpha was provided.
• Note: set_alpha() must be called with single scalar, otherwise a TypeError exception is raised.
• Note: it seems that both the scatter constructor and set_alpha writes alpha into the color code because of the following
  • set_alpha(None) does not restore alpha to 1.0, but reverts the negative override on depthscale
  • set_alpha(1.0) overrides the A channel of RGBA colors
• Remark: the above behavior is inconsistent with line plotting for the following reasons:
  • there, set_alpha(None) reverts only the explicitly provided alpha, not the A channel of RGBA colors
  • there, set_alpha(1.0) does not overwrite the A channel of RGBA colors, but the minimum of the alpha and A is used, however I would prefer product over minimum.

fig = plt.figure()
ax = fig.gca(projection = "3d")
sc1 = ax.scatter([0,1], [0,1], [1,1], color = (1,0,0,1), alpha=0.3, depthshade = True)
sc2 = ax.scatter([0,1], [0,1], [2,2], color = (1,0,0,1), alpha=0.3, depthshade = True)
sc3 = ax.scatter([0,1], [0,1], [3,3], color = (1,0,0,0.3), depthshade = True)
sc4 = ax.scatter([0,1], [0,1], [4,4], color = (1,0,0,0.3), depthshade = True)
sc1.set_alpha(1.0)
sc2.set_alpha(None)
sc3.set_alpha(None)
sc4.set_alpha(1.0)
ax.set_zticks([1,2,3,4])
ax.set_zticklabels(['a=.3->1', 'a=.3->N', 'A=.3->N', 'A=.3->1'])
fig.show()

Matplotlib version

• Operating system: Windows
• Matplotlib version: 3.0.2
• Matplotlib backend: both module://ipykernel.pylab.backend_inline and Qt5
• Python version: 3.6.7 |Anaconda custom (64-bit)| (default, Dec 10 2018, 20:35:02) [MSC v.1915 64 bit (AMD64)]