__contains__, overlaps, overlap_rate, intersection, get_area methods provide errors/wrong results for geos AreaDef / SwathDef

[ghiggi]

Problem description

The methods __contains__, overlaps, overlap_rate, intersection, contains, get_area implemented in BaseDefinition class employs a polygon defined by the centroid of the pixel corners of the SwathDef / AreaDef to perform the geometrical computations.
While this works correctly for many classical AreaDefinition, for:

• Geostationary AreaDef all methods crash (because of Inf values at the corners) (see also outer_boundary_corners and __contains__ raise ValueError on some areas #393, AreaDefinition.__contains__ raises ValueError #268) 3outer_boundary_corners and __contains__ raise ValueError on some areas #393
• SwathDefinition, the corners of lats/lons arrays does not provide meaningful values in the case of ascending and descending orbits within SwathDefinition ...

Code example for GEO AreaDef

import satpy
geometry_def = satpy.resample.get_area_def("goes_east_abi_f_1km")
geometry_def.corners   # --> Illegal (lon, lat) coordinates: (inf, inf)
geometry_def.outer_boundary_corners # --> Illegal (lon, lat) coordinates: (inf, inf)
geometry_def.get_area() # --> Illegal (lon, lat) coordinates: (inf, inf)
geometry_def.overlaps(geometry_def) # --> Illegal (lon, lat) coordinates: (inf, inf)
geometry_def.overlap_rate(geometry_def) # --> Illegal (lon, lat) coordinates: (inf, inf)
geometry_def.intersection(geometry_def) # --> Illegal (lon, lat) coordinates: (inf, inf)
(10,20) in geometry_def # __contains__ method  # --> Illegal (lon, lat) coordinates: (inf, inf)

Proposed solution:

• Definition of boundary method for SwathDefinition and AreaDefinition to retrieve AreaBoundary class
• Definition of SphPolygon from AreaBoundary using self.polygon()
• Enhancement of spherical.py geometrical operations
• Upgrade of BaseDefinition methods to use self.boundary().polygon().<method>(...)

Draft PR coming in the next days ...

[gerritholl]

In your "see also" you referred once to #268 and twice to #393. Was there a third issue you meant to link?