Handle empty geometry about geopandas HOT 8 CLOSED

@fscottfoti I think rather than skipping the geometries, you'd include them with None for the geometry column.

Note that GeoPandas does not currently handle the null geometry case at the moment i.e., df.area would raise an exception, where it should probably return nan for rows with null geometries. I'll file a separate issue.

Going forward, you would use df[df['geometry'].notnull()].to_file(...) to filter out null geometries and write the result. You shouldn't have to do this if certain file types can handle features with no geometries.

from geopandas.

See #139. Slight change from above, set missing geometries as np.nan.

from geopandas.

Makes perfect sense. I think you should be able to write empty geometries too, just because to_file should be the inverse of from_file.

from geopandas.

I agree, but only if the output format can handle missing geometries. If Fiona handles it, we should make sure we're doing that right. I don't know enough about all the formats to know how common this is. Just read the GeoJSON spec and found out that it's acceptable for Features to have null geometry, never knew that before.

But if the output requires a geometry to be present, and you pass in null ones, then I think it should crash. It's up to the user to pass in the data properly.

from geopandas.

Makes sense. Yes, I hadn't thought much before about missing geometries either, then I got some degenerate data from a client today. If I get a chance I will see if I can fix things up to read/write missing data. Thx.

from geopandas.

There is a method to fill null values in a GeoSeries, fillna(), that will by default write empty polygons to null values. That doesn't do you any good if the read is failing, but it does give you a consistent geometry for shapely operations.

from geopandas.

Awesome - thank you the fix!

from geopandas.

Fixed by #138

from geopandas.