make mesh2d_make_global in line with interacter behaviour

[veenstrajelmer]

Is your feature request related to a problem? Please describe.
When generating a global grid with mesh2d_make_global, the resulting grid is sort of flipped with respect to the reference in interacter. This is inconvenient, since it complicates conversions in other tools like the xugrid.to_nonperiodic() conversion.

To reproduce:

from meshkernel import DeleteMeshOption, MeshKernel, ProjectionType, GeometryList
import matplotlib.pyplot as plt
plt.close('all')
import xugrid as xu
import numpy as np

mk = MeshKernel(projection=ProjectionType.SPHERICAL)
mk.mesh2d_make_global(num_longitude_nodes=192, num_latitude_nodes=100)

line_array = np.array([[-68.37701613,  21.48864852],
       [-71.26512097,  51.0113413 ],
       [  1.65952621,  57.30306271],
       [ 15.37802419,  19.55273423],
       [-32.99773185, -14.32576568],
       [-62.60080645,  11.32509854],
       [-68.37701613,  21.48864852]])
xx = line_array[:,0]
yy = line_array[:,1]
delete_pol_geom = GeometryList(x_coordinates=xx, y_coordinates=yy)
mk.mesh2d_delete(geometry_list=delete_pol_geom, 
                 delete_option=DeleteMeshOption.INSIDE_NOT_INTERSECTED,
                 invert_deletion=False)

mesh2d = mk.mesh2d_get()

fig, ax = plt.subplots()
mesh2d.plot_edges(ax)
ax.plot(xx, yy, 'r-')

Gives:

Describe the solution you'd like
Get a flipped version of the figure above, in other words: get the continuous longitude line at lon=-180 instead of lon=180.

Additional context
Additionally, the algorithm seems not to be robust for different lat/lon values.

When converting it to non-periodic with xugrid we do not see the around-the-back lines anymore, which is expected. This example is not self-contained, it requires also the code block above.

# xugrid conversion
xu_grid = xu.Ugrid2d.from_meshkernel(mesh=mesh2d, projected=False, crs=4326)
# TODO: this to_nonperiodic conversion now only partly works, it does not result in grid in the upper/lower left corners if the figure.
xu_grid_flat = xu_grid.to_nonperiodic(xmax=180)

#fig, ax = plt.subplots()
#xu_grid.plot()
fig, ax = plt.subplots()
xu_grid_flat.plot()

Gives:

When supplying a low number of longitude values, we get orphan edges at the min/max latitudes

from meshkernel import MeshKernel, ProjectionType
import matplotlib.pyplot as plt
plt.close('all')
mk = MeshKernel(projection=ProjectionType.SPHERICAL)
mk.mesh2d_make_global(num_longitude_nodes=20, num_latitude_nodes=100)
mesh2d = mk.mesh2d_get()
fig, ax = plt.subplots()
mesh2d.plot_edges(ax)

When supplying a low number of latitude values, the resulting grid does not extent to +/- 90 latitudes:

from meshkernel import MeshKernel, ProjectionType
import matplotlib.pyplot as plt
plt.close('all')
mk = MeshKernel(projection=ProjectionType.SPHERICAL)
mk.mesh2d_make_global(num_longitude_nodes=100, num_latitude_nodes=20)
mesh2d = mk.mesh2d_get()
fig, ax = plt.subplots()
mesh2d.plot_edges(ax)

[veenstrajelmer]

Meshkernel 6 flipped the grid left/right, which is good. However, the around-the-back edges are now also incorrectly removed. The orphan edges are fixed. The extent from -90 to +90 is not fixed