Javier,
The Conda vs your build difference can be easily explained because Conda
sets PROJ_NETWORK=ON in the conda environment.
When PROJ_NETWORK=ON, you can also get the expected result by adding
--bbox as well to cs2cs.
Now, the real issue is why PROJ_NETWORK=ON without --box does
(negatively) affect results .. ? The reason is that in that situation
the heuristics first finds the NKG related operations, and does not
think it needs to do extra efforts to consider intermediate CRSs to get
other solutions. Unfortunately the NKG solutions only cover a small
proportion of the area of use of the source/target CRS, so they are not
good general purpose solutions.
I've tuned this in https://github.com/OSGeo/PROJ/pull/4244
Now:
$ echo 46 6 0 2024.6 | PROJ_NETWORK=OFF bin/cs2cs EPSG:9989 EPSG:4937 -d 9
45.999994986 5.999991132 -0.001870790 2024.6
$ echo 46 6 0 2024.6 | PROJ_NETWORK=ON bin/cs2cs EPSG:9989 EPSG:4937 -d 9
45.999994986 5.999991132 -0.001870790 2024.6
$ echo 46 6 0 2024.6 | PROJ_NETWORK=OFF bin/cs2cs EPSG:9989 EPSG:4937
-d 9 --bbox 5.5,45.5,6.5,46.5
45.999994986 5.999991132 -0.001870790 2024.6
$ echo 46 6 0 2024.6 | PROJ_NETWORK=ON bin/cs2cs EPSG:9989 EPSG:4937 -d
9 --bbox 5.5,45.5,6.5,46.5
45.999994986 5.999991132 -0.001870790 2024.6
(Anybody wanting to do a "Fast, reliable and geodetic-grade qualiy
coordinate operation inference from a geodetic database" master or PhD
thesis ... ? I may consider mentoring them, unless they decide to put
some AI in the equation ;-) ;-). Or perhaps that's actually the way to
do it. Spy all conversations occurring in offices of geodetic
institutions and coordinate transformation software editors,
speech-to-text them, ingest that into a LLM, and add some sanity filter
on the garbage it spits out)
Even
--
http://www.spatialys.com
My software is free, but my time generally not.
_______________________________________________
PROJ mailing list
[email protected]
https://lists.osgeo.org/mailman/listinfo/proj
