WIP: Update 6 baseline images due to GMT 6.1.0 changes

[seisman]

This PR updates the baseline images of 6 failing tests and removes the xfail markers.

These tests fail due to slight changes between GMT 6.0 and GMT 6.1, and are not related to grid data.

Reminders

• Run make format and make check to make sure the code follows the style guide.
• Add tests for new features or tests that would have caught the bug that you're fixing.
• Add new public functions/methods/classes to doc/api/index.rst.
• Write detailed docstrings for all functions/methods.
• If adding new functionality, add an example to docstrings or tutorials.

[weiji14]

A bit hesitant to approve this but ok.

We should look into using https://matplotlib.org/3.2.2/api/testing_api.html#matplotlib.testing.decorators.check_figures_equal for the grd* changes, i.e. check that PyGMT is reproducing the images output by the current GMT version, otherwise this is a never-ending game.

[seisman]

OK. We can hold this PR for a while.

[seisman]

We should look into using https://matplotlib.org/3.2.2/api/testing_api.html#matplotlib.testing.decorators.check_figures_equal for the grd* changes, i.e. check that PyGMT is reproducing the images output by the current GMT version, otherwise this is a never-ending game.

I'm a little confused here. When I first read the link, I thought we can avoid unnecessary image updates by increasing the RMS threshold, but I don't see why check_figures_equal is preferred over image_comparison.

One more thing, we are using the pytest-mpl plugin, but the link above is for matplotlib.

[weiji14]

I'm a little confused here. When I first read the link, I thought we can avoid unnecessary image updates by increasing the RMS threshold, but I don't see why check_figures_equal is preferred over image_comparison.

One more thing, we are using the pytest-mpl plugin, but the link above is for matplotlib.

Sorry for the confusion. I wanted to point out how matplotlib states that

"This [check_figures_equal] decorator should be preferred over image_comparison when possible in order to keep the size of the test suite from ballooning.".

This is a totally different way of testing, different to using pytest-mpl. I.e. Instead of testing against a static baseline image, I'm considering whether we should just check the differences between 1. a plot generated using xarray.DataArray grids and 2. NetCDF file grids. The risk here is that there's a bug which affects both xarray.DataArray and NetCDF grids, but currently all our problems stem from using xarray.DataArray grids. If we assume that NetCDF grids are always fine in GMT, then this would be safe to do.

[seisman]

check the differences between 1. a plot generated using xarray.DataArray grids and 2. NetCDF file grids.

It sounds a really good idea.

[seisman]

We should look into using https://matplotlib.org/3.2.2/api/testing_api.html#matplotlib.testing.decorators.check_figures_equal for the grd* changes, i.e. check that PyGMT is reproducing the images output by the current GMT version, otherwise this is a never-ending game.

Tried a little bit of the mpl decorator. It doesn't work for us, because it has some matplotlib-specific codes. But the idea is great. I think we need to implement our own way to compare two figures, like:

fig1 = pygmt.Figure()
fig1.basemap(....)

fig2 = pygmt.Figure()
fig2.basemap(....)

pygmt.check_figure_equal(fig1, fig2, tol="0.01")