Remove expected failures on grdview tests #589
Conversation
The xarray based slicing isn't as precise as grdcut, so some images were a few pixels off when plotted in 3D.
pygmt/tests/test_grdview.py
Outdated
| @check_figures_equal() | ||
| def test_grdview_with_perspective_and_zaxis_frame(gridfile, grid): | ||
| """ | ||
| Run grdview by passing in a grid and plotting an annotated vertical | ||
| z-axis frame. | ||
| """ | ||
| fig = Figure() | ||
| fig.grdview(grid=grid, perspective=[225, 30], zscale=0.005, frame="zaf") | ||
| return fig | ||
| fig_ref = Figure() | ||
| fig_ref.grdview(grid=gridfile, perspective=[225, 30], zscale=0.005, frame="zaf") | ||
| fig_test = Figure() | ||
| fig_test.grdview(grid=grid, perspective=[225, 30], zscale=0.005, frame="zaf") | ||
| return fig_ref, fig_test |
There was a problem hiding this comment.
This test_grdview_with_perspective_and_zaxis_frame is a very strange one. It passes when the full test suite is ran using make test, but was failing locally for me everytime I ran pytest --mpl --verbose --doctest-modules pygmt/tests/test_grdview.py. Even weirder is that the problem seems to be on the NetCDF plot rather than the xarray one (now I've seen everything)! The difference is a doubling on the x and y-axis frame, will report this to upstream GMT in a bi (Edit: it's at GenericMappingTools/gmt#4181):
| NetCDF (expected, but wrong) | xarray.DataArray (test, correct?!!) |
|---|---|
 |
 |
Exact error is:
matplotlib.testing.exceptions.ImageComparisonFailure: Image sizes do not match expected size: (1534, 2664, 3) actual size (1521, 2638, 3)
There was a problem hiding this comment.
The xarray.DataArray result is "incorrect", because GMT takes the input xarray grid as a Cartesian grid, although grid.gmt.gtype=1.
There was a problem hiding this comment.
Ok, so there's several layers of 'wrong-ness' here it seems? The NetCDF grdview plot is wrong because it has a double frame (without the zebra stripes), and the xarray grid is wrong because it's using the Cartesian plain map frame? But these are 'absolute' wrongs i.e. bugs, and our new @check_figures_equal decorator only does a 'relative' comparison to ensure we're reproducing GMT plots in PyGMT (be it right or wrong).
What I don't understand is why the tests pass when ran using make test (false negative), but fails with pytest --mpl --verbose --doctest-modules pygmt/tests/test_grdview.py. Seems like another flaky test issue.
|
Just two suggestions:
fig = [None, None]
for i, grid in enumerate((gridfile, xgrid)):
fig[i] = Figure()
fig[i].grdview(grid=gridfile, perspective=[135, 15])
return fig
# return fig[0], fig[1] |
weiji14
changed the title
Otherwise GMT (fig_ref) might plot double-lined x and y axis frame, while PyGMT (fig_test) plots a single-lined frame.
Final round of tests to migrate to DVC for the `grdview` plotting module. This is related to #1131, and is almost like reverting #589 actually. * Migrate test_grdview baseline images to dvc * Update fig.grdview baseline images for GMT 6.2.0rc1 * Plot fancy frame for test_grdview_with_perspective * Update fig.grdview baseline images for GMT 6.2.0rc2 * Use projection="Q15c+" for test_grdview_with_perspective
Final round of tests to migrate to DVC for the `grdview` plotting module. This is related to GenericMappingTools#1131, and is almost like reverting GenericMappingTools#589 actually. * Migrate test_grdview baseline images to dvc * Update fig.grdview baseline images for GMT 6.2.0rc1 * Plot fancy frame for test_grdview_with_perspective * Update fig.grdview baseline images for GMT 6.2.0rc2 * Use projection="Q15c+" for test_grdview_with_perspective
Description of proposed changes
Removes the expected failures (xfail) pytest marks used as a workaround in #503 to deal with
grdviewfailures.Part 1 to fixing #451. Part 2 will be to deal with xfails on
makecptand othergrd*modules.Reminders
make formatandmake checkto make sure the code follows the style guide.doc/api/index.rst.