Geometry path issues in geostationary projection with geometry.geoms

[TheoMathurin]

geoviews 1.12.0
holoviews 1.18.3
bokeh 3.4.1
cartopy 0.23.0
geos 3.12.1
shapely 2.0.4

With coastline.geoms objects in geostationary projection, paths display weird segments extending apparently infinitely towards the top-left corner.

Not sure this is due to geoviews per se though. I've tried various environments including with older versions of all these libraries and they now all seem to exhibit this behaviour while it used to be fine. I've made sure to delete local natural earth geometries cached by cartopy to fetch the latest (could it come from there?)

Note that it does not happen with a bare gv.feature.coastline, but only with its .geoms (which I use with resample_geometry).

import panel as pn
import geoviews as gv
import cartopy.crs as ccrs

gv.extension('bokeh')


plots = [gv.feature.coastline.geoms(res).options(projection=ccrs.Geostationary()) for res in ['110m', '50m', '10m']]

pn.Row(*plots)

No error is thrown in python or javascript.

[ahuang11]

Potential workaround:

import panel as pn
import geoviews as gv
import cartopy.crs as ccrs

gv.extension("bokeh")


plots = [
    gv.Points((0, 0), crs=ccrs.Geostationary()).opts(projection=ccrs.Geostationary())
    * gv.project(gv.feature.coastline().geoms(res), projection=ccrs.Geostationary())
    for res in ["110m", "50m", "10m"]
]

pn.Row(*plots)

[TheoMathurin]

Many thanks @ahuang11 for the idea.

In my admittedly complex use case wrapping geoms into gv.project somehow results in the geometry not showing. Can't really pinpoint a cause right now though.

[TheoMathurin]

Does anyone know where this issue might come from? I'd like to fix this as otherwise the resampled geometry cannot be used.

[ahuang11]

To pin it down, I'd look at what gv.project does vs the crs/projection kwargs.

I'd also check if you can reproduce in vanilla matplotlib + cartotpy.