[PROJ] Relevant descrepancies between Sinusoidal and Homolosine

[Even Rouault]

On mardi 10 mars 2020 12:42:15 CET Luí­s Moreira de Sousa wrote:
> Hi Even,
> 
> I have not yet checked the formulas from that source, but regarding the
> current Homolosine implementation, its outputs are the expected. So perhaps
> it is applying the spherical Sinusoidal and the ellipsoidal Mollweide?

No, before my PR, that would be the reverse: it would apply the ellipsoidal 
Sinusoidal and the (spherical, but there's only spherical implemented) 
Mollweide. igh doesn't do real computations by itself: it just dispatches to 
the existing sinu or moll methods depending on the input coordinate.

Maybe, a proper fix would probably be add an ellipsoidal formulation to 
Mollweide. But igh would then be much more complicated because the dispatching 
logic strongly relies on the fact that there's an identity between northing 
and latitude from what I can see.

So I guess making sure igh uses spherical formulas for both sinuosidal and 
mollweide, as I intented in my PR, is probably better than the current 
situation where there's a mix ?

-- 
Spatialys - Geospatial professional services
http://www.spatialys.com