[GRASS-dev] Fwd: [PROJ] PROJ 6RC2

[Markus Metz]

On Sat, Feb 23, 2019 at 10:26 PM Even Rouault <even.rouault at spatialys.com>
wrote:
>
> On samedi 23 février 2019 15:59:28 CET Markus Neteler wrote:
> > Hi devs,
> >
> > there is some discussion about packages that need to be patched to use
> > proj_api.h or proj.h which will no longer be available in recent PROJ.
> >
> > Is there anything to do in GRASS GIS (trunk)?
>
> From what I can see in
> https://github.com/GRASS-GIS/grass-ci/blob/master/lib/proj/do_proj.c
> Markus Metz made use of proj.h API of PROJ 5. This should still work with
PROJ
> 6.
>
> However the way of building pipelines in
> https://github.com/GRASS-GIS/grass-ci/blob/master/lib/proj/do_proj.c#L92
> from the PROJ string of the source and target CRS (PROJ strings are not a
good
> way to capture a CRS definition, mostly due to losing the original datum
> information) will be suboptimal in PROJ 6 since still using a
early-binding
> approach with WGS84 as a pivot, instead of the new late-binding
capabilities
> brought by PROJ 6.
> If GRASS stores EPSG code and/or WKT in addition/instead of PROJ strings,
then
> using proj_create_crs_to_crs() will lead to more accurace results in
> circumstances where WGS84 is not a natural pivot, like the case of
> NAD27+NVGD29 to NAD83+NAVD88 that Paul Ramsey exposed in
> http://blog.cleverelephant.ca/2019/02/proj4-postgis.html

I have seen the magic lookup you implemented in proj_create_crs_to_crs()
for PROJ 6, really impressive! It's on my TODO list to make use of
proj_create_crs_to_crs() in GRASS (for PROJ 5 it was not worth the effort
because proj_create_crs_to_crs() constructed a very simple pipeline).

GRASS already stores EPSG code if plrovided which could be enhanced to
store any SRID as supported by proj's +init=..., according to the init
files coming with RPOJ (https://proj4.org/resource_files.html#init-files).

We could also store WKT, but then I would suggest to enable this only for
PROJ6+ and store only WKT2, not WKT1.

Markus M
>
> Even
>
> --
> Spatialys - Geospatial professional services
> http://www.spatialys.com
> _______________________________________________
> grass-dev mailing list
> grass-dev at lists.osgeo.org
> https://lists.osgeo.org/mailman/listinfo/grass-dev
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.osgeo.org/pipermail/grass-dev/attachments/20190224/ab065fb0/attachment.html>