good catch! Do you happen to have a snippet of code that could be used as a small test?

Snippet of code to reproduce fixed results (bugged on live)

from mpl_toolkits.mplot3d import Axes3D
from mpl_toolkits.mplot3d.art3d import Poly3DCollection
import matplotlib.pyplot as plt

fig = plt.figure()
ax = Axes3D(fig)
x = [0, 1, 1, 0]
y = [0, 0, 1, 1]
z = [0, 1, 0, 1]
verts = [list(zip(x, y, z))]

alpha=0.5
fc = "C0"

pc = Poly3DCollection(verts, alpha = alpha, facecolors=fc, linewidths=1)

ax.add_collection3d(pc)
plt.show()

I am not sure that is this the right fix as if you apply alpha multiple times they will stack (ex setting the alpha to 0.5 twice will effectively be 0.25).

The same fix should probably be applied to edge colors as well.

Hmm I'll look into that soon, and yeah this seems applicable to edgecolors as well. I just kept it at facecolors for simplicity since I'm a first-time contributer.

Alpha stacking doesn't seem to be an issue from what I see. I've tested this by applying alpha (setting alpha to 0.5) 6 times (effectively reducing down to 0.01 if stacking occurs), but it remained 0.5, demonstrating that stacking doesn't happen. Let me know if I'm missing something (I'm uploading this response in a bit of a rush!).

Code for reproduction:

from mpl_toolkits.mplot3d import Axes3D
from mpl_toolkits.mplot3d.art3d import Poly3DCollection
import matplotlib.pyplot as plt

fig = plt.figure()
ax = Axes3D(fig)
x = [0, 1, 1, 0]
y = [0, 0, 1, 1]
z = [0, 1, 0, 1]
verts = [list(zip(x, y, z))]

alpha=0.5
fc = "C0"

pc = Poly3DCollection(verts, alpha = alpha, facecolors=fc, linewidths=1)
pc.set_alpha(alpha)
pc.set_alpha(alpha)
pc.set_alpha(alpha)
pc.set_alpha(alpha)
pc.set_alpha(alpha)
ax.add_collection3d(pc)
plt.show()

Thanks @DanielMatu This could use a test so this doesn't get missed again, if you are up for it: I think in tests/test_mplot3d. Bonus points if it doesn't require a new image comparison.

Sounds good. I'll get on that asap.

What's that travis error that prevents this from passing the tests? It doesn't seem to be part of this PR?

forced a rebuild

Would alpha need to be applied to edgecolor as well? (see #9559)

Would alpha need to be applied to edgecolor as well?

I just kept it at facecolors for simplicity since I'm a first-time contributer.

+1 for also applying this to edgecolors, as otherwise, this PR won't fully fix the parent issue.

What's that travis error that prevents this from passing the tests? It doesn't seem to be part of this PR?

Travis CI fails for Mac OS with /usr/bin/python: No module named pip possibly due to travis-ci/travis-ci#8829 . This doesn't happen on master, maybe syncing with it could help?

Looking at the original issue, there were a few different ways to get the right result and a few ways to get the wrong result. It might make sense to test a few different ways to make sure that this works right.

ping @DanielMatu Any interest in still working on this?

Ummm, my apologies, somehow I broke this by force pushing....