[BUG] Fix alpha bug on 3D PathCollection plots. #25478

ricmperes · 2023-03-16T13:17:41Z

PR Summary

Quick fix of bug on treatment of alpha parameter on the mplot3d toolkit, reported in issue #25446, following @oscargus and @tacaswell directions on the issue thread.

For Patches, no changed seemed to be needed. Might need further testing since it's the same piece of code repeated:

matplotlib/lib/mpl_toolkits/mplot3d/art3d.py

Lines 588 to 596 in 63b9b04

    
           def _maybe_depth_shade_and_sort_colors(self, color_array): 
        
               color_array = ( 
        
                   _zalpha(color_array, self._vzs) 
        
                   if self._vzs is not None and self._depthshade 
        
                   else color_array 
        
               ) 
        
               if len(color_array) > 1: 
        
                   color_array = color_array[self._z_markers_idx] 
        
               return mcolors.to_rgba_array(color_array, self._alpha)

A script to test the fix can be found in this gist: https://gist.github.com/ricmperes/a69d876d2cf7333ae9cf713ecb25b474

Result after fix:

PR Checklist

Documentation and Tests

Has pytest style unit tests (and pytest passes)
Documentation is sphinx and numpydoc compliant (the docs should build without error).
New plotting related features are documented with examples.

Release Notes

New features are marked with a .. versionadded:: directive in the docstring and documented in doc/users/next_whats_new/
API changes are marked with a .. versionchanged:: directive in the docstring and documented in doc/api/next_api_changes/
Release notes conform with instructions in next_whats_new/README.rst or next_api_changes/README.rst

github-actions

Thank you for opening your first PR into Matplotlib!

If you have not heard from us in a while, please feel free to ping @matplotlib/developers or anyone who has commented on the PR. Most of our reviewers are volunteers and sometimes things fall through the cracks.

You can also join us on gitter for real-time discussion.

For details on testing, writing docs, and our review process, please see the developer guide

We strive to be a welcoming and open project. Please follow our Code of Conduct.

oscargus · 2023-03-16T13:50:47Z

Thanks! Can you please add a test for this as well?

I think that plotting something like in the original issue and then checking that the alpha-value of the "correct" position(s) (row 10000 to 19999) of scat3d.get_facecolor() is 0 would do. (No need for image tests.)

ksunden · 2023-03-16T16:26:26Z

Test failure is related:

The test:

matplotlib/lib/mpl_toolkits/mplot3d/tests/test_axes3d.py

Lines 441 to 494 in 63b9b04

    
           @pytest.mark.parametrize('depthshade', [True, False]) 
        
           @check_figures_equal(extensions=['png']) 
        
           def test_scatter3d_sorting(fig_ref, fig_test, depthshade): 
        
               """Test that marker properties are correctly sorted.""" 
        
               y, x = np.mgrid[:10, :10] 
        
               z = np.arange(x.size).reshape(x.shape) 
        
               sizes = np.full(z.shape, 25) 
        
               sizes[0::2, 0::2] = 100 
        
               sizes[1::2, 1::2] = 100 
        
               facecolors = np.full(z.shape, 'C0') 
        
               facecolors[:5, :5] = 'C1' 
        
               facecolors[6:, :4] = 'C2' 
        
               facecolors[6:, 6:] = 'C3' 
        
               edgecolors = np.full(z.shape, 'C4') 
        
               edgecolors[1:5, 1:5] = 'C5' 
        
               edgecolors[5:9, 1:5] = 'C6' 
        
               edgecolors[5:9, 5:9] = 'C7' 
        
               linewidths = np.full(z.shape, 2) 
        
               linewidths[0::2, 0::2] = 5 
        
               linewidths[1::2, 1::2] = 5 
        
               x, y, z, sizes, facecolors, edgecolors, linewidths = [ 
        
                   a.flatten() 
        
                   for a in [x, y, z, sizes, facecolors, edgecolors, linewidths] 
        
               ] 
        
               ax_ref = fig_ref.add_subplot(projection='3d') 
        
               sets = (np.unique(a) for a in [sizes, facecolors, edgecolors, linewidths]) 
        
               for s, fc, ec, lw in itertools.product(*sets): 
        
                   subset = ( 
        
                       (sizes != s) | 
        
                       (facecolors != fc) | 
        
                       (edgecolors != ec) | 
        
                       (linewidths != lw) 
        
                   ) 
        
                   subset = np.ma.masked_array(z, subset, dtype=float) 
        
                   # When depth shading is disabled, the colors are passed through as 
        
                   # single-item lists; this triggers single path optimization. The 
        
                   # following reshaping is a hack to disable that, since the optimization 
        
                   # would not occur for the full scatter which has multiple colors. 
        
                   fc = np.repeat(fc, sum(~subset.mask)) 
        
                   ax_ref.scatter(x, y, subset, s=s, fc=fc, ec=ec, lw=lw, alpha=1, 
        
                                  depthshade=depthshade) 
        
               ax_test = fig_test.add_subplot(projection='3d') 
        
               ax_test.scatter(x, y, z, s=sizes, fc=facecolors, ec=edgecolors, 
        
                               lw=linewidths, alpha=1, depthshade=depthshade)

Looks like an interaction with depthshade = True. In this case it's a "compare equal" not a "compare to static reference" test, so pretty clear that the two methods of plotting should result in the same image (plotting in subsets of markers with the same properties vs plotting in one call).

tacaswell · 2023-03-16T19:01:04Z

lib/mpl_toolkits/mplot3d/art3d.py

-        return mcolors.to_rgba_array(color_array, self._alpha)
+
+        if self._alpha is not None:
+            converted_alphas = color_array[:, 3] * self._alpha


Are we sure this will never get an RGB (vs RGBA) array?

Hi @tacaswell. Good point, if depthshape is specified as False a RGBA type array might get passed. I put an escape clause for this case.

Suggested change

converted_alphas = color_array[:, 3] * self._alpha

return mcolors.to_rgba_array(color_array, self._alpha)

ricmperes · 2023-06-30T19:30:12Z

Debugging why the test_scatter3d_sorting is failing I ran it with a clean version of matplotlib and the output seems to have no depth-related shading, likely because alpha=1 is parsed. To be sure, when depthshade=True this is not the intended behaviour, right?
Output bellow:

scottshambaugh · 2024-07-16T16:55:32Z

Since depthshade=True is the default, I would expect that manually setting the alpha parameter would override the depth shading, but I don't actually know what the intent here is. @ricmperes I know this PR has been hanging for a bit - are you still interested in working on it?

scottshambaugh · 2025-01-30T16:50:37Z

Hi @ricmperes, a fix to scatter plot depth shading was just merged in here: #29287

I think that should resolve the test failure you were seeing. Are you still interested in working on this PR?

ricmperes · 2025-01-31T12:10:16Z

Perfect, that's great! I can rebase, test and if fixed, then close the PR. Thanks!

scottshambaugh · 2025-04-12T17:19:42Z

Hi @ricmperes, are you still interested in working on this?

ricmperes added 2 commits March 16, 2023 14:01

Fix alpha bug on 3D PathCollection plots.

9b5a0ab

Fix flake8 linting errors

a941214

github-actions bot reviewed Mar 16, 2023

View reviewed changes

oscargus added the topic: mplot3d label Mar 16, 2023

oscargus linked an issue Mar 16, 2023 that may be closed by this pull request

[Bug]: Nan values in scatter 3d plot show in black colour when alpha parameter is passed. #25446

Open

tacaswell reviewed Mar 16, 2023

View reviewed changes

scottshambaugh added the PR: bugfix Pull requests that fix identified bugs label Mar 17, 2023

jklymak added the status: needs revision label Jun 14, 2023

ricmperes added 3 commits June 30, 2023 18:53

Handle RGB instead of RGBA case.

09d2179

Clean main if clause.

69ad517

Merge branch 'matplotlib:main' into main

47de4ba

github-actions bot added the status: needs rebase label Jan 30, 2025

scottshambaugh added the status: orphaned PR label Jan 30, 2025

scottshambaugh removed status: needs revision status: orphaned PR labels Jan 31, 2025

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

[BUG] Fix alpha bug on 3D PathCollection plots. #25478

[BUG] Fix alpha bug on 3D PathCollection plots. #25478

ricmperes commented Mar 16, 2023

github-actions bot left a comment

oscargus commented Mar 16, 2023

ksunden commented Mar 16, 2023

tacaswell Mar 16, 2023

ricmperes Jun 30, 2023

ricmperes commented Jun 30, 2023

scottshambaugh commented Jul 16, 2024

scottshambaugh commented Jan 30, 2025

ricmperes commented Jan 31, 2025

scottshambaugh commented Apr 12, 2025

	def _maybe_depth_shade_and_sort_colors(self, color_array):
	color_array = (
	_zalpha(color_array, self._vzs)
	if self._vzs is not None and self._depthshade
	else color_array
	)
	if len(color_array) > 1:
	color_array = color_array[self._z_markers_idx]
	return mcolors.to_rgba_array(color_array, self._alpha)

	converted_alphas = color_array[:, 3] * self._alpha
	return mcolors.to_rgba_array(color_array, self._alpha)

[BUG] Fix alpha bug on 3D PathCollection plots. #25478

Are you sure you want to change the base?

[BUG] Fix alpha bug on 3D PathCollection plots. #25478

Conversation

ricmperes commented Mar 16, 2023

PR Summary

PR Checklist

github-actions bot left a comment

Choose a reason for hiding this comment

oscargus commented Mar 16, 2023

ksunden commented Mar 16, 2023

tacaswell Mar 16, 2023

Choose a reason for hiding this comment

ricmperes Jun 30, 2023

Choose a reason for hiding this comment

ricmperes commented Jun 30, 2023

scottshambaugh commented Jul 16, 2024

scottshambaugh commented Jan 30, 2025

ricmperes commented Jan 31, 2025

scottshambaugh commented Apr 12, 2025