Extent of axes dependent of order of points being shown

[LucaMarconato]

Changing the order of what is being rendered may alter the extent being shown in the plot.

Here below is not a way to reproduce the bug because it would be laborious, but it shows an example of what is going on. This code

        import matplotlib.pyplot as plt
        import spatialdata_plot

        # otherwise pre-commits will remove the "unused" import
        _ = spatialdata_plot

        ax = plt.gca()
        sdata.pl.render_shapes(element="by_polygons", na_color=(1.0, 0.7, 0.7, 0.5)).pl.show(ax=ax)
        sdata.pl.render_shapes(element="by_circles", na_color=(1.0, 0.7, 0.7, 0.5)).pl.show(ax=ax)
        sdata.pl.render_shapes(element="values_polygons", color="categorical_in_obs").pl.show(ax=ax)
        sdata.pl.render_shapes(element="values_circles", color="categorical_in_obs").pl.show(ax=ax)
        sdata.pl.render_points(color="categorical_in_ddf", size=10.0, palette="tab10").pl.show(ax=ax)
        # sdata.pl.render_shapes(element="values_circles", color="categorical_in_obs").pl.show(ax=ax) # <- this line is identical to the one 2 lines above!
        plt.show()

gives this plot:

If we uncomment the last line, we get this plot instead:

If we plot only the line that is commented in the code above, this is shown:

Expected behaviors:

1. the final extent should not dependent of the order of elements being plot
2. the expected extent is the screenshots above doesn't fit the actual extent of the data

[LucaMarconato]

Note: I didn't look at the extent code but I saw a few PRs modifying it. If the problem arises because of edges cases of coordinate transformations I am happy to help refactoring the get_extent() code inside SpatialData.

[LucaMarconato]

More details: I think the problem is in render_points().

[timtreis]

Linking scverse/spatialdata#288

[timtreis]

Note from call with @LucaMarconato: When axs object is passed, check for current axes limit and, if needed, extend them but not overwrite.

[timtreis]

@LucaMarconato can you try to reproduce your error?