[GRASSLIST:10643] Re: r.proj takes a very long time / does not complete

[Hamish]

Hi Paul,

> > That would be very handy if r.proj could estimate the best output
> > resolution as well. I remember there was some discussion about
> > completely replacing r.proj warper with gdal warper. Any news?
> 
> The way I see it is that they perform two different, complementary 
> functions. gdalwarp is useful if you have a whole map/image you want
> to  convert to a new projection and not lose any of the data. r.proj
> is useful  if you already have a region of interest you're working on
> in the GIS, and  want to import only the portion of the larger
> map/image that covers this region, at the same resolution and with the
> same boundaries etc. as your  existing maps in your location.
> 
> Does that make sense?

(again, sorry for very late responses)


I don't know if this goes for r.proj method=nearest,bilinear,cubic as 
well, but for i.rectify the idea was to use the GDAL warp *API*, not 
the gdalwarp program.

GRASS and GDAL use the same base code for 1st,2nd,3rd order transforms
(AFAICT); both have the same bad bugs if you try 3rd order, e.g..

The difference is that the GDAL version is better maintained and
optimized, plus has nifty features like a thin plate spline option.

As we already depend on the gdal libraries, I think it's a good idea
to let someone else maintain that bit of the code.
see  https://intevation.de/rt/webrt?serial_num=2952


Hamish