Test not working as intended

The test test_image.test_image_cursor_formatting() is intended to test mouseover formatting of masked image data.
see issue #15103 and resolution in PR #15140 .

def test_image_cursor_formatting():
    fig, ax = plt.subplots()
    im = ax.imshow(np.zeros((4, 4)))

    data = np.ma.masked_array([0], mask=[True])
    assert im.format_cursor_data(data) == '[]'
    ...

However, this test does not work as intended, because it Artist.format_cursor_data() will in normal circumstances call _ColorizerInterface._format_cursor_data_override():

    def format_cursor_data(self, data):
        if np.ndim(data) == 0 and hasattr(self, "_format_cursor_data_override"):
            return self._format_cursor_data_override(data)
        ...

However, because of the way test_image_cursor_formatting() is implemented the array sent to im.format_cursor_data() is 1d, and _format_cursor_data_override() is therefore not called.

i.e. this test does not test the formatting of masked image data as intended.

The relevant behavior has another test

The relevant test (test_image.test_image_cursor_formatting()) is 6 years old. Since then, a more comprehensive test for im.format_cursor_data() has been developed, reproduced below.

@pytest.mark.parametrize(
    "data, text", [
        ([[10001, 10000]], "[10001.000]"),
        ([[.123, .987]], "[0.123]"),
        ([[np.nan, 1, 2]], "[]"),
        ([[1, 1+1e-15]], "[1.0000000000000000]"),
        ([[-1, -1]], "[-1.0]"),
        ([[0, 0]], "[0.00]"),
    ])
def test_format_cursor_data(data, text):
    from matplotlib.backend_bases import MouseEvent

    fig, ax = plt.subplots()
    im = ax.imshow(data)

    xdisp, ydisp = ax.transData.transform([0, 0])
    event = MouseEvent('motion_notify_event', fig.canvas, xdisp, ydisp)
    assert im.format_cursor_data(im.get_cursor_data(event)) == text

This tests im.format_cursor_data() with masked values (nan), and I therefore believe it is not necessary to adapt test_image.test_image_cursor_formatting() so that it tests the relevant behavior, and im.format_cursor_data() can instead be removed.

Why did this attract my attention

I noticed this because I am adapting _format_cursor_data_override() to work with multivariate data, see #29876 and related PRs, and need to develop new tests for this behavior. I attempted to base my tests on test_image.test_image_cursor_formatting() but this proved futile, and I will instead implement tests based on test_image.test_format_cursor_data().