Thanks for your contribution!

I think it will be beneficial to have a What's new entry for sure.

What is required though is a test for this. I am not sure how it should work exactly, but at least adding a PatchCollection to an Axes, adding a legend and make sure that there is an entry in the legend is a start. Probably one should also check that the marker/patch is correct as well as the label.

The test can either go in https://github.com/matplotlib/matplotlib/blob/main/lib/matplotlib/tests/test_collections.py or https://github.com/matplotlib/matplotlib/blob/main/lib/matplotlib/tests/test_legend.py (not sure which is correct, one can argue for both).

Hint 1: Here is the description how to add a what's new entry: https://github.com/matplotlib/matplotlib/tree/main/doc/users/next_whats_new

Hint 2: This is about the test code we need:

import matplotlib.pyplot as plt
from matplotlib.collections import PatchCollection
from matplotlib.patches import Circle, Rectangle
import matplotlib.colors as mcolors

fig, ax = plt.subplots()
pc = PatchCollection([
        Circle((0, 0), radius=1, facecolor='red', edgecolor='green', linewidth=3, linestyle='--'),
        Rectangle((0.5, 0.5), 1, 1),
    ],
    match_original=True,
    label='my collection')
ax.add_collection(pc)

handles, labels = ax.get_legend_handles_labels()
assert len(labels) == 1
assert labels[0] == 'my_collection'
assert mcolors.same_color(handles[0].get_facecolor(), 'red')
assert mcolors.same_color(handles[0].get_edgecolor(), 'green')
assert handles[0].get_linewidth() == 3
assert handles[0].get_linestyle() == '--'

Thank you for your help and tips @oscargus @timhoffm

I will try to finish this today!

Edit: I will handle the linting failures and push again.

Hi,

I have pushed some new changes. Now, both facecolors and edgecolors are handled. I also added a what's_new document, as well as a test in test_legend.py. However, I have a few problems/questions.

1.) All the tests are successful, with the exception of the linestyle. Even though the linestyle is what I expected (visibly), the assertion fails. This is because for some reason, the output of "handles[0].get_linestyle()[0]" is (0.0, [11.100000000000001, 4.800000000000001]). Hence, when compared to "--" (which is supposed to be equivalent to (0, (5, 5))) if I'm not wrong), the assertion fails. (This is illustrated in the attached picture)

2.) Is it possible to change the shape of the object inside the legend box? Even though the first object in the collection is a circle, it appears as a rectangle in the legend box, which is clearly not ideal (also shown in the attached picture). Is there an attribute that controls this? Is there any get_xxx method that could help me here?

The dash pattern is converted to numbers here

So it’s configurable, but assuming it matches the lines.dashed_pattern shown in the default matplotlibrc would give (0, (3.7, 1.6)).

Then there is also some scaling where that pattern is multiplied by the linewidth, which is obviously 3 in your case.

So (11.1, 4.8) seems right to me.

I'd say that a PatchCollection would be identified with color rather than shape, so it should not be a problem that it shows up using a rectangle. (Compare to #23914 where probably we do not want the Eiffel tower to show up in a legend...)

Regarding linestyle, PatchCollection (or maybe even collections in general) do return the dash pattern rather than the actual linestyle ('--'/'dashed'). This is sorted out in #23056 but that is not merged yet. I think the best way is to use something like np.allclose([11.1, 4.8], ...) as there is some sort of floating-point error from the scaling.

Oops. Will fix the test soon and make final changes to finish the PR.

Just checking in on the status here? It seems almost complete.

@FarukFS are you still interested in working on this one?

I think it would be good to get this over the line, so I took the liberty of rebasing and fixing the test. The only difference is that the test now checks the handle linestyle against the collection linestyle instead of specific numbers (which will depend on the rcParams).

Removed some trailing whitespace and added the handler in the pyi file. Hopefully now CI is happy...

Hmmm. Seems this wasn’t as close to done as I’d assumed.

Any idea why none of the new code appears to have coverage, even though there's a new test?

At least some of that was because legend didn't get called in the test, so create_artists didn't get called.

I think this one was ready to go but, since I pushed to it, I do not think I should approve it. If it's not ready I think it should be marked as orphaned.