The previous broken cases now work well with this branch.

Thanks, taking of WIP. The old fix was a wack-a-mole fix. Fixed one thing and broke another. No way to fix both without this rewrite. Once tested with PyGMT and approved I will merge and then work on similar upgrades for a few other places in the API where this appears, and it may likely lead to improvements such as doing -o2,1,2,2 etc being allowed.

Thanks. And at the end we'll need another release.

The original contour test now works, but there is still a failure for rose.

The load_fractures_compilation loads the @fractures_06.txt file and returns a DataArray with two columns, "length" and "azimuth". In the first figure, "length" and "azimuth" is passed, so the figure is correct. In the second figure, "azimuth" and "length" is passed, so incols=[1, 0] is used to swap the input columns, but the second figure is wrong:

import pygmt
from pygmt.datasets import load_fractures_compilation

data = load_fractures_compilation()

data1 = data[["length", "azimuth"]]
fig = pygmt.Figure()
fig.rose(
    data=data1,
    region=[0, 1, 0, 360],
    sector=10,
    diameter="10c",
    frame=["x0.2g0.2", "y30g30", "+glightgray"],
    color="red3",
    pen="1p",
    orientation=False,
    norm=True,
    vectors=True,
    no_scale=True,
    shift=False,
)
fig.savefig("rose1.png")

data2 = data[["azimuth", "length"]]
fig = pygmt.Figure()
fig.rose(
    data=data2,
    incols=[1, 0],
    region=[0, 1, 0, 360],
    sector=10,
    diameter="10c",
    frame=["x0.2g0.2", "y30g30", "+glightgray"],
    color="red3",
    pen="1p",
    orientation=False,
    norm=True,
    vectors=True,
    no_scale=True,
    shift=False,
)
fig.savefig("rose2.png")

First figure	Second figure

OK, will look at this later - have some errands that need to be done first...

Should be OK now, @seisman.

Now all PyGMT tests pass.

@PaulWessel , you mentioned "Only testing matrices and vectors here as testing datasets revealed some unrelated issues that will be dealt with later". I'm debugging GenericMappingTools/pygmt#1440 and find that the col_pos used in:

is incorrect. Do you know if this is related to the known issues for for gmtapi_import_dataset? If not, I can continue debugging but don't want to repeat efforts.

No, please continue, and let me know if you need help.