[FIXED] Error converting to GeoDataFrame with `np.nan` geometries

Issue

I have a pandas DataFrame that I am trying to convert to a GeoDataframe. I have some geometries that are np.nan values and I get an error when attempting to convert them.

import pandas as pd
import geopandas as gpd 
import numpy as np
from shapely import wkt

df = pd.DataFrame({
                   id: [0, 1, 2, 3],
                   geo: [
                         'POLYGON ((35.067912 -106.700884, 35.066714 -106.70128, 35.066555 -106.700697, 35.066492 -106.700468, 35.066134 -106.700367, 35.066088 -106.700129, 35.066062 -106.700005, 35.067519 -106.699412, 35.06763 -106.69937, 35.067659 -106.699537, 35.067737 -106.699977, 35.067814 -106.700378, 35.067891 -106.700773, 35.067912 -106.700884))', 
                         np.nan,
                         'POLYGON ((35.067912 -106.700884, 35.066714 -106.70128, 35.066555 -106.700697, 35.066492 -106.700468, 35.066134 -106.700367, 35.066088 -106.700129, 35.066062 -106.700005, 35.067519 -106.699412, 35.06763 -106.69937, 35.067659 -106.699537, 35.067737 -106.699977, 35.067814 -106.700378, 35.067891 -106.700773, 35.067912 -106.700884))',
                         np.nan
                        ]
                  })

   usgeo = gpd.GeoDataFrame(df)

   usgeo['geometry'] = usgeo['geom'].apply(wkt.loads)


/Applications/Anaconda/anaconda3/lib/python3.9/site-packages/shapely/geos.py in read(self, text)
    325         """Returns geometry from WKT"""
    326         if not isinstance(text, str):
--> 327             raise TypeError("Only str is accepted.")
    328         text = text.encode()
    329         c_string = c_char_p(text)

TypeError: Only str is accepted.

How do I convert to geoDataFrame with empty / np.nan as geometries? I'd prefer not to drop the rows as I care about the other information in the dataframe.

Solution

You can't parse np.nan as wkt. You can however handle this in your apply:

usgeo['geometry'] = usgeo['geom'].apply(
    lambda x: wkt.loads(x) if isinstance(x, str) else None
)

See the geopandas docs section on missing and empty geometries for a discussion of the differences between missing values (None or np.nan) and empty geometries (e.g. shapely.geometry.GeometryCollection()). These two have different interpretations in geopandas.

Answered By - Michael Delgado