geoda.cars geometry column not automatic

[kylebarron]

Unclear if you consider this an issue or not, but it would be nice if gpd.read_file(geodatasets.get_path()) always constructed a dataframe with a geometry column. For the geoda.cars dataset, the geometry column is full of None values, but there are lat/lon columns:

gdf = gpd.read_file(geodatasets.get_path("geoda.cars"))

Is the right solution:

• Fix this under the hood somehow
• Tell the user to construct a geometry column from the lat/lon columns after loading
• Inform the user about the underlying CSV structure to pass the right OGR open options

[martinfleis]

The empty geometry column seems to be issue only for Fiona engine. Reading it with engine="pyogrio" (the new default in the geopandas 1.0), it is not there, but you also get a pandas.DataFrame (that is a bit of inconsistency we need to resolve somehow).

About the solutions, I am not sure what to do.

geodatasets is just giving you an unaltered file and I believe it should stay that way.

The rest is related to a better documentation on a dataset level, which I am not sure how would I do. Maybe via a note in the dataset json?

There is also another option and that is a removal of the dataset if we think that gpd.read_file(geodatasets.get_path("name")) should always result in a GeoDataFrame in an expected form.

Since the goal of geodatasets is to provide mostly educational and illustrational data, having a CSV with point coordinates feels useful but there was never a deep discussion on the scope, so happy to hear your thoughts.