Raise warning if both c and facecolors are used in scatter plot (... and related improvements in the test suite). #29130

NGWi · 2024-11-12T23:18:22Z

This is to fulfill @tacaswell's recommendation (at the PyData sprint I joined last week) that for a quick fix to close issue #24404, we raise a warning when such a script is attempted.
The complaint in that issue was: that if both c and facecolors args are specified for a plot's markers (specifically a scatterplot), then the facecolors arg is ignored, while a user might expect that facecolors would be used for the fill and the 'c' color would remain at the edges.
I have the code raise a warning in the style of others I saw in _axes.py :
"You passed both c and facecolor/facecolors for the markers. "
"Matplotlib is ignoring the facecolor "
"in favor of what you specified in c. "
"This behavior may change in the future."
I am hoping that in the future we can implement c as edgecolors in that case, and we are not forced at some point to cause the plot to fail.

PR checklist

[✅ ] "Resolves [Bug]: Cannot use empty markers in scatter #24404 " is in the body of the PR description to Resolve #24404
[✅ ] new and changed code is tested
[N/A] Plotting related features are demonstrated in an example
[I think N/A, but let me know.] New Features and API Changes are noted with a directive and release note
[I think N/A, but let me know. ] Documentation complies with general and docstring guidelines

…he future we may implement c as edgecolors in that case or cause plot to fail.

github-actions

Thank you for opening your first PR into Matplotlib!

If you have not heard from us in a week or so, please leave a new comment below and that should bring it to our attention. 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.

NGWi · 2024-11-12T23:29:02Z

Would you like me to push another commit and PR that removes the offending extra line and whitespace (flake8 fail) or edit the PR?
Update: I edited the PR-staged code.

NGWi · 2024-11-13T01:34:52Z

I see the other failures are because of two tests failing: one because my warning was triggered for test_legend_pathcollection_labelcolor_markfacecolor_cmap

        fig, ax = plt.subplots()
        facecolors = mpl.cm.viridis(np.random.rand(10))
        ax.scatter(
            np.arange(10),
            np.arange(10),
            label='#1',
            c=np.arange(10),
            facecolor=facecolors
        )

And the other because another warning was triggered.
That second one has been plaguing PRs since this morning: https://github.com/matplotlib/matplotlib/actions/workflows/tests.yml

=========================== short test summary info ============================
FAILED lib/matplotlib/tests/test_backends_interactive.py::test_interactive_backend[toolmanager-MPLBACKEND=wxagg-BACKEND_DEPS=wx] - Failed: Subprocess failed to test intended behavior
<frozen _collections_abc>:982: UserWarning: Treat the new Tool classes introduced in v1.5 as experimental for now; the API and rcParam may change in future versions.
Traceback (most recent call last):
  File "<string>", line 1, in <module>
  File "/home/runner/work/matplotlib/matplotlib/lib/matplotlib/tests/test_backends_interactive.py", line 232, in _test_interactive_impl
    assert result.getvalue() == result_after.getvalue()
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
AssertionError

I just raised issue #29131 about that one.
But I'll have to analyze the first test to see if it's failing only because of my intentional improvement in behavior.

To pass flake8 formatting check

timhoffm · 2024-11-13T13:33:15Z

lib/matplotlib/axes/_axes.py

+                "You passed both c and facecolor/facecolors for the markers. "
+                "Matplotlib is ignoring the facecolor "
+                "in favor of what you specified in c. "


Slightly more concise:

Suggested change

"You passed both c and facecolor/facecolors for the markers. "

"Matplotlib is ignoring the facecolor "

"in favor of what you specified in c. "

"Passing both c and facecolor/facecolors for the markers "

"is ambiguous. facecolor is ignored in in favor of c. "

Thank you for reviewing the comment. Your change is more concise, and I understand the argument that matplotlib developers consider passing both as ambiguous.
However, a user intentionally passing both will likely not understand why it is ambiguous. They probably assumed incorrectly, as the issue's Original Poster did, that c defines both a face color and edge color and that just like adding edge color works, the face color can also be overwritten.

timhoffm · 2024-11-13T13:35:16Z

Would you like me to push another commit and PR that removes the offending extra line and whitespace (flake8 fail) or edit the PR? Update: I edited the PR-staged code.

Yes, you've don this correctly: Please update this PR and do not create new ones on the same topic. You can choose whether to add additional commits or force-push a new version.

…otlib into issue24404_addwarning

NGWi · 2024-11-13T15:44:50Z

I tried out the test code that was failing
test_legend_pathcollection_labelcolor_markfacecolor_cmap

        fig, ax = plt.subplots()
        facecolors = mpl.cm.viridis(np.random.rand(10))
        ax.scatter(
            np.arange(10),
            np.arange(10),
            label='#1',
            c=np.arange(10),
            facecolor=facecolors
        )

As I suspected, facecolors doesn't do anything in that case, so my addition justifiably raises the warning. This, of course, alters the outcome from the assertion's previous expectation.
I will change the test to:

        fig, ax = plt.subplots()
        ax.scatter(
            np.arange(10),
            np.arange(10),
            label='#1',
            c=np.arange(10)
        )

which produces the indentical plot as before (I tested it with .canvas.tostring_rgb()).
However, I imagine that will not update all your online test suites. I hope you will approve my PR regardless.

NGWi · 2024-11-13T16:09:22Z

That failing test was added in Commit 66ec3a0 .
I see the intention was to test if the label color in the legend is correctly set when using a colormap.
So I, instead, am only dropping the c argument in the test:

    fig, ax = plt.subplots()
    facecolors = mpl.cm.viridis(np.random.rand(10))
    ax.scatter(
        np.arange(10),
        np.arange(10),
        label='#1',
        facecolor=facecolors
    )

    leg = ax.legend(labelcolor='markerfacecolor')
    for text, color in zip(leg.get_texts(), ['k']):
        assert mpl.colors.same_color(text.get_color(), color)

This will let the random color assignment shine through, but it won't raise a warning. The only problem would be if a visual compare test were done with its old output.

lib/matplotlib/tests/test_legend.py

NGWi · 2024-11-17T15:34:31Z

I want to mention that I am working on an alternative branch where I add an ec (edge colors) arg to .scatter() that parallels c's functionality and would come before the current 'edgecolors' arg (which I bump to kwargs). This would make it explicit that c should not be used for specifically setting edge colors and return intuitive, symmetrical functionality to .scatter().

(timhoffm: accidental edit reverted)

NGWi · 2024-11-18T14:32:38Z

@timhoffm, your edit to my comment above is full of good points.
I would still want to explore if I can add the functionality of having two independent colormap objects running in parallel, one for face and one for edge. Don't worry, I am treading extra carefully as I am aware that I am new here, plus I am proposing a much bigger shift than what @tacaswell already proposed.

timhoffm · 2024-11-18T14:48:15Z

@timhoffm, your edit to my comment above is full of good points.

Sorry for editing your comment. I intended to click "quote reply" but accidentally hit the wrong button 🐑. I've reverted the edit.

For reference, here is my comment:

I'm sceptical on this. c is primarily used for color mapping. With the current implementation of ScalarMappable you cannot have two independent color mappings (one for face and one for edge). Therefore, I don't see how ec could be parallel to c. Note also that ec already exists as an alias for edgecolors.

timhoffm · 2024-11-18T14:53:41Z

I would still want to explore if I can add the functionality of having two independent colormap objects running in parallel.

You should look into #28658 and #28428, which generalize the colormapping concept. The case of separate edgecolor mapping is not explicitly contained, but fits into that context.

…hoffm

…hoffm's points.

NGWi · 2024-11-19T19:59:34Z

Codecov doesn't like that in my test I wrote:

+     def get_next_color():
+         return 'blue'

I think it considers the output of that function as untested. But I was just following the lead of three tests above it, which it is also complaining about. Should I change my test, or ignore it?

The other failing check is build time taking over an hour (including the test suite, I think). I'm not sure that is because of something I did, though.

timhoffm · 2024-11-20T10:02:07Z

lib/matplotlib/tests/test_axes.py

-        def get_next_color():
-            return 'blue'  # currently unused
-


Let's not move the get_next_color() functions out of the tests. This produces an indirection (harder to read) and implicit coupling of the tests. There's also value in the "currently unused" comment in here.

Compared to this, the repetition of two lines is clearly the lesser evil.

Even better would be a local function

def get_next_color(): raise AssertionError("Color cycle should not be accessed")

I have no problem rolling back those commits.
But what do you want to do about the fact that CodeCov raises an "Uncovered" flag by every incidence of return 'blue'? It even failed my code because I had added another incidence of that: (https://github.com/matplotlib/matplotlib/pull/29130/checks?check_run_id=33222717934)

CodeCov does not strictly need to pass, but we have it for information only.

Rolled back.

…plified return.

story645 · 2024-12-04T01:23:18Z

lib/matplotlib/tests/test_axes.py

@@ -3027,7 +3027,7 @@ def get_next_color():
     ])
 def test_parse_scatter_color_args_edgecolors(kwargs, expected_edgecolors):
    def get_next_color():


Suggested change

def get_next_color():

def get_next_color(): # pragma no cover

Pretty sure it has to go next to the function declaration because you're telling coverage to ignore this whole block

You're right. That was my instinct as well, but I saw @jefromyers did it next to the return statement

matplotlib/lib/matplotlib/tests/test_axes.py

Line 8298 in 04ce3e9

return 'blue' # pragma: no cover

(line 9076 in the current iteration).
I just corrected it.

NGWi · 2024-12-04T02:35:43Z

The AppVeyor check failed, but after reading the error logs, and doing the same workflow on my Mac, I believe it's a Windows problem unrelated to my change.

…he future we may implement c as edgecolors in that case or cause plot to fail.

…hoffm

…hoffm's points.

…plified return.

NGWi · 2024-12-04T02:47:15Z

I rebased my branch but incidentally created duplicate commits on top. I should have used interactive rebase, sorry.

…fro's test.

NGWi · 2024-12-04T04:20:36Z

Is there anything else I can do to improve my Pull Request?

lib/matplotlib/tests/test_legend.py

Co-authored-by: Tim Hoffmann <[email protected]>

NGWi · 2024-12-04T16:24:14Z

Anything else?

NGWi · 2024-12-05T01:45:15Z

Thank you, @timhoffm

story645 · 2025-01-03T14:35:51Z

Thanks @NGWi, sorry it took so long to follow up, hope to see more contributions from you! 🥳

NGWi · 2025-01-03T15:32:17Z

Thanks 😀

…

On Fri, Jan 3, 2025 at 9:36 AM hannah ***@***.***> wrote: Thanks @NGWi <https://github.com/NGWi>, sorry it took so long to follow up, hope to see more contributions from you! 🥳 — Reply to this email directly, view it on GitHub <#29130 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ANEGD73HTFOXV3LFGJ4K4O32I2N55AVCNFSM6AAAAABRVFGIR6VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDKNRZGMYTINJUGM> . You are receiving this because you were mentioned.Message ID: ***@***.***>

QuLogic · 2025-01-07T00:00:36Z

@story645 story645 added this to the v3.10.1 milestone Jan 3, 2025

You didn't trigger a backport, if this milestone was intentional.

story645 · 2025-01-07T04:49:38Z

Uch what's the command again? I think this should get back ported since it's just an error check, right?

timhoffm · 2025-01-07T09:06:16Z

@meeseeksdev backport to v3.10.x

…are used in scatter plot (... and related improvements in the test suite).

…130-on-v3.10.x Backport PR #29130 on branch v3.10.x (Raise warning if both c and facecolors are used in scatter plot (... and related improvements in the test suite).)

Raise warning if both c and facecolors are used in scatter plot. In t…

5a26bb9

…he future we may implement c as edgecolors in that case or cause plot to fail.

github-actions bot reviewed Nov 12, 2024

View reviewed changes

Corrected flake8 formatting warnings.

f1966c2

Removed offending whitespace and extra line.

53a0825

To pass flake8 formatting check

timhoffm reviewed Nov 13, 2024

View reviewed changes

Merge branch 'issue24404_addwarning' of https://github.com/NGWi/matpl…

a777a15

…otlib into issue24404_addwarning

Corrected old test that qualified for the new warning.

8031809

rcomer reviewed Nov 16, 2024

View reviewed changes

lib/matplotlib/tests/test_legend.py Show resolved Hide resolved

NGWi added 4 commits November 18, 2024 11:58

Made message more concise, and straightened old test, to fulfill @tim…

8ae213c

…hoffm's points.

Added test.

49f03f7

Corrected flake8 linting objections.

8234066

Added direct test of _parse method to attempt to make Codecov happy.

38c092a

timhoffm reviewed Nov 20, 2024

View reviewed changes

NGWi force-pushed the issue24404_addwarning branch from 611a9dc to 38c092a Compare November 20, 2024 21:17

Added pragma tags for code coverage to ignore unused but untested sim…

e6e49aa

…plified return.

story645 reviewed Dec 4, 2024

View reviewed changes

NGWi added 3 commits December 3, 2024 21:41

Raise warning if both c and facecolors are used in scatter plot. In t…

39e1899

…he future we may implement c as edgecolors in that case or cause plot to fail.

Corrected flake8 formatting warnings.

06c8a76

Corrected old test that qualified for the new warning.

72138c6

NGWi added 5 commits December 3, 2024 21:41

Made message more concise, and straightened old test, to fulfill @tim…

319b92b

…hoffm's points.

Added test.

e348dfc

Corrected flake8 linting objections.

fcf6ce2

Added direct test of _parse method to attempt to make Codecov happy.

9601975

Added pragma tags for code coverage to ignore unused but untested sim…

a18364c

…plified return.

NGWi force-pushed the issue24404_addwarning branch from e6e49aa to a18364c Compare December 4, 2024 02:45

Corrected placement of pragma tags and added 'currently unused' to Je…

310a15f

…fro's test.

timhoffm reviewed Dec 4, 2024

View reviewed changes

lib/matplotlib/tests/test_legend.py Outdated Show resolved Hide resolved

NGWi and others added 2 commits December 4, 2024 10:11

Tim's suggestion for naming test's list of colors.

7c01a7b

Co-authored-by: Tim Hoffmann <[email protected]>

Corrected call to changed variable name.

fd3d9db

NGWi changed the title ~~Raise warning if both c and facecolors are used in scatter plot.~~ Raise warning if both c and facecolors are used in scatter plot (... and related improvements in the test suite). Dec 4, 2024

timhoffm approved these changes Dec 5, 2024

View reviewed changes

story645 merged commit df49afc into matplotlib:main Jan 3, 2025
39 checks passed

story645 added this to the v3.10.1 milestone Jan 3, 2025

meeseeksmachine pushed a commit to meeseeksmachine/matplotlib that referenced this pull request Jan 7, 2025

Backport PR matplotlib#29130: Raise warning if both c and facecolors …

5357f0d

…are used in scatter plot (... and related improvements in the test suite).

meeseeksmachine mentioned this pull request Jan 7, 2025

Backport PR #29130 on branch v3.10.x (Raise warning if both c and facecolors are used in scatter plot (... and related improvements in the test suite).) #29423

Merged

claytharrison mentioned this pull request Feb 19, 2025

[Bug]: Cannot use empty markers in scatter #24404

Open

ksunden mentioned this pull request Mar 3, 2025

merge up v3.10.1 #29700

Merged

5 tasks

	def get_next_color():
	def get_next_color(): # pragma no cover

Uh oh!

Raise warning if both c and facecolors are used in scatter plot (... and related improvements in the test suite). #29130

Raise warning if both c and facecolors are used in scatter plot (... and related improvements in the test suite). #29130

Uh oh!

Conversation

NGWi commented Nov 12, 2024

PR checklist

Uh oh!

github-actions bot left a comment

Choose a reason for hiding this comment

Uh oh!

NGWi commented Nov 12, 2024 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

NGWi commented Nov 13, 2024 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

timhoffm Nov 13, 2024

Choose a reason for hiding this comment

Uh oh!

NGWi Nov 13, 2024

Choose a reason for hiding this comment

Uh oh!

timhoffm commented Nov 13, 2024

Uh oh!

NGWi commented Nov 13, 2024

Uh oh!

NGWi commented Nov 13, 2024

Uh oh!

Uh oh!

NGWi commented Nov 17, 2024 • edited by timhoffm Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

NGWi commented Nov 18, 2024

Uh oh!

timhoffm commented Nov 18, 2024

Uh oh!

timhoffm commented Nov 18, 2024

Uh oh!

NGWi commented Nov 19, 2024

Uh oh!

timhoffm Nov 20, 2024 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

NGWi Nov 20, 2024

Choose a reason for hiding this comment

Uh oh!

rcomer Nov 20, 2024

Choose a reason for hiding this comment

Uh oh!

NGWi Nov 20, 2024

Choose a reason for hiding this comment

Uh oh!

story645 Dec 4, 2024

Choose a reason for hiding this comment

Uh oh!

NGWi Dec 4, 2024

Choose a reason for hiding this comment

Uh oh!

NGWi commented Dec 4, 2024

Uh oh!

NGWi commented Dec 4, 2024 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

NGWi commented Dec 4, 2024

Uh oh!

Uh oh!

NGWi commented Dec 4, 2024

Uh oh!

NGWi commented Dec 5, 2024

Uh oh!

Uh oh!

story645 commented Jan 3, 2025

Uh oh!

NGWi commented Jan 3, 2025 via email

Uh oh!

QuLogic commented Jan 7, 2025

Uh oh!

story645 commented Jan 7, 2025

NGWi commented Nov 12, 2024 •

edited

Loading

NGWi commented Nov 13, 2024 •

edited

Loading

NGWi commented Nov 17, 2024 •

edited by timhoffm

Loading

timhoffm Nov 20, 2024 •

edited

Loading

NGWi commented Dec 4, 2024 •

edited

Loading