[PROJ] Proposal for a geodetic deformation model format

[Even Rouault]

Hi Jochem,

(top-posting, because, for some reason, part of the list traffic doesn't reach 
my email box...)

> 2. In a NTv2 grid shift file with a subgrid, discontinuities could occur at 
the edge between the parent grid and the subgrid.
> Maybe PROJ. already does interpolate in that way? I didn’t check the source 
code.

No PROJ does it the easy way: if the point to correct falls in a (sub)grid, it 
interpolates only within that subgrid.
Actually, this is slightly more subtle than that: PROJ selects a subgrid, even 
if the point to transform is slightly outside it (epsilon based on a 1 / 10000 
of the (sub)grid resolution: see
https://github.com/OSGeo/PROJ/blob/master/src/apply_gridshift.cpp#L127 )
And for the reverse transformation, it also stays within the subgrid it 
initially selected.

Even

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