[gdal-dev] +proj=wintri +over option works weel in gdalwarp but not in ogr2ogr

[Even Rouault]

Joaquim,

I've looked at that. Experimentally, it doesn't seem that the current 
maths for wintri in +over mode give reasonable results for longitudes > 
360deg in absolute value (and I'm not sure if there would be a natural 
extension possible. Far beyond my knowledge of the underlying maths).

gdalwarp does a little better by default because it couples the forward 
and inverse transformer to check the domain of validity and 
experimentally must find this limitation and will thus give up for 
source pixels whose longitude > 360 deg. For ogr2ogr / vector this is 
more complicated as reprojection of a single geometry  (here cl540.gpkg 
is a single polygon with longitudes in [-540,540] range). is a "all 
points succeed to reproject => OK" / "one single point fails to 
reproject => KO" type of situation, at least in the logic followed by 
OGRGeometry::transform() logic.

So while the following patch to PROJ is probably technically correct, it 
wouldn't help in practice

diff --git a/src/projections/aitoff.cpp b/src/projections/aitoff.cpp
index 9d060999f..cce319157 100644
--- a/src/projections/aitoff.cpp
+++ b/src/projections/aitoff.cpp
@@ -59,6 +59,11 @@ static PJ_XY aitoff_s_forward(PJ_LP lp, PJ *P) { /* 
Spheroidal, forward */
      struct pj_opaque *Q = static_cast<struct pj_opaque *>(P->opaque);
      double c, d;

+    if( fabs(lp.lam) > 2 * M_PI ) {
+        proj_errno_set(
+                    P, PROJ_ERR_COORD_TRANSFM_OUTSIDE_PROJECTION_DOMAIN);
+        return xy;
+    }
      c = 0.5 * lp.lam;
      d = acos(cos(lp.phi) * cos(c));
      if (d != 0.0) { /* basic Aitoff */

The only solution I can offer is that you must explictly clip your 
source geometry to the validity range of wintri with -clipsrc -360 -90 
360 90.

Even


Le 30/04/2023 à 03:11, Joaquim Manuel Freire Luís via gdal-dev a écrit :
>
> Even, this is a kind of continuation of the subject that I brought up 
> in https://github.com/OSGeo/gdal/issues/7644
>
> The file size limitation of attachments doesn’t help to make this case 
> easily reproducible, but the story is, I can make that rectangular map 
> using gdalwarp and ogr2ogr that use the +over option and +proj=vandg
>
> (see 
> https://forum.generic-mapping-tools.org/t/best-projection-for-rectangular-world-map/3715/82?u=joaquim)
>
> But when I try the same with +proj=wintry the gdalwarp op works well 
> but the ogr2ogr doesn’t.
>
> (see 
> https://forum.generic-mapping-tools.org/t/best-projection-for-rectangular-world-map/3715/85?u=joaquim)
>
> You can more less reproduce this case using this file
>
> http://fct-gmt.ualg.pt/tmp/cl540.gpkg
>
> ogr2ogr  -t_srs "+proj=wintri +over" cl540_wintri.gpkg cl540.gpkg
>
> Why would that be in this case?
>
> Joaquim
>
>
> _______________________________________________
> gdal-dev mailing list
> gdal-dev at lists.osgeo.org
> https://lists.osgeo.org/mailman/listinfo/gdal-dev

-- 
http://www.spatialys.com
My software is free, but my time generally not.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.osgeo.org/pipermail/gdal-dev/attachments/20230430/5942fff7/attachment.htm>