CI: matplotlib failures

[TomAugspurger]

https://travis-ci.org/pandas-dev/pandas/jobs/542921706#L2423

=================================== FAILURES ===================================
_______________________ TestDataFramePlots.test_subplots _______________________
[gw1] linux -- Python 3.6.8 /home/travis/miniconda3/envs/pandas-dev/bin/python
self = <pandas.tests.plotting.test_frame.TestDataFramePlots object at 0x7f507ed05128>
    @pytest.mark.slow
    def test_subplots(self):
        df = DataFrame(np.random.rand(10, 3),
                       index=list(string.ascii_letters[:10]))
    
        for kind in ['bar', 'barh', 'line', 'area']:
            axes = df.plot(kind=kind, subplots=True, sharex=True, legend=True)
            self._check_axes_shape(axes, axes_num=3, layout=(3, 1))
            assert axes.shape == (3, )
    
            for ax, column in zip(axes, df.columns):
                self._check_legend_labels(ax,
                                          labels=[pprint_thing(column)])
    
            for ax in axes[:-2]:
                self._check_visible(ax.xaxis)  # xaxis must be visible for grid
                self._check_visible(ax.get_xticklabels(), visible=False)
                self._check_visible(
>                   ax.get_xticklabels(minor=True), visible=False)
pandas/tests/plotting/test_frame.py:387: 
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ 
self = <pandas.tests.plotting.test_frame.TestDataFramePlots object at 0x7f507ed05128>
collections = <a list of 2 Text xticklabel objects>, visible = False
    def _check_visible(self, collections, visible=True):
        """
        Check each artist is visible or not
    
        Parameters
        ----------
        collections : matplotlib Artist or its list-like
            target Artist or its list or collection
        visible : bool
            expected visibility
        """
        from matplotlib.collections import Collection
        if not isinstance(collections,
                          Collection) and not is_list_like(collections):
            collections = [collections]
    
        for patch in collections:
>           assert patch.get_visible() == visible
E           AssertionError
pandas/tests/plotting/common.py:137: AssertionError

I'm looking into it now, but will have to go in 30 minutes.

[TomAugspurger]

Seems to be a change in matplotlib. The test passes for me locally, with 3.0.3, fails with 3.1.0. (cc @tacaswell, in case this wasn't deliberate. Sorry I don't have an MPL-only reproducer right now)

Just going to change the assert for 3.1.0+ for now.

[simonjayhawkins]

@TomAugspurger i looked into this, xref #26582, the best i could come up with is that we are using an undocumented method in the test. maybe not part of the official api. our code suggests that we assume a patch object but are testing a text object.

[tacaswell]

My first guess was https://matplotlib.org/api/api_changes.html?highlight=api#major-minor-tick-collisions where we now hide ticks that collide by default, but the issue is things are visible that should not be.

Does the image actually look different?

[TomAugspurger]

No it looks the same afaict.

…

On Jun 7, 2019, at 18:45, Thomas A Caswell ***@***.***> wrote: My first guess was https://matplotlib.org/api/api_changes.html?highlight=api#major-minor-tick-collisions where we now hide ticks that collide by default, but the issue is things are visible that should not be. Does the image actually look different? — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub, or mute the thread.