[gdal-dev] Strange (expected?) behaviour exporting multipart geometry to ESRI Shapefile format

[Andrew C Aitchison]

On Wed, 16 Feb 2022, Even Rouault wrote:

> That said I've tried replicated using 
> https://github.com/qgis/QGIS/files/8058606/Issue47288_Polygons_GPKG.zip from 
> the QGIS ticket and converting it to shapefile. The issue here is that the 
> original geometry is not (considered as) valid:
>
> $ ogrinfo Issue47288_Polygons.gpkg -sql "select st_isvalid(geom) from 
> Issue47288_Polygons" -al -q
> GEOS warning: Self-intersection at or near point 352107.37400937825 
> 5662263.0605139844 54.963999999999999
>
> Layer name: SELECT
> OGRFeature(SELECT):0
>   st_isvalid(geom) (Integer) = 0

> As far as I can see in this multipolygon, there are 2 issues: one of the part 
> is repeated, and the 2 "small" rectangular polygons that share the 
> 352107.37400937825 5662263.0605139844 54.963999999999999 points are likely 
> non-planar.

> My inclination would be to consider that in the case where we have a
> multipolygon with non -coplanar parts we should probably be avoiding
> considering the ones that are non-coplanar are inner/outer rings of
> others.
>
> I've ticketed that in https://github.com/OSGeo/gdal/issues/5315

IIUC, whether a polygon is planar depends upon the CRS, just as
transforming a straight line to another CRS may produce a non-straight
line.
If so, do we have to worry about this ?

-- 
Andrew C. Aitchison					Kendal, UK
 			andrew at aitchison.me.uk