Folium: ValueError: key_on `'FIPS'` not found

Question

I'm working on a Folium choropleth map, using a shapefile from the Census bureau and religion data from the US Religion Census. I can get both sets of data loaded and cleaned up, so I have two DFs:

geojson

<class 'geopandas.geodataframe.GeoDataFrame'>
RangeIndex: 3234 entries, 0 to 3233
Data columns (total 2 columns):
 #   Column    Non-Null Count  Dtype   
---  ------    --------------  -----   
 0   FIPS      3234 non-null   object  
 1   geometry  3234 non-null   geometry

map_data

<class 'pandas.core.frame.DataFrame'>
RangeIndex: 3142 entries, 0 to 3141
Data columns (total 2 columns):
 #   Column     Non-Null Count  Dtype  
---  ------     --------------  -----  
 0   FIPS       3142 non-null   object 
 1   PctAllPop  3142 non-null   float64

When I try to create the map with the following code, I get a value error:

m1 = folium.Map(location=[40, -96], zoom_start=4,tiles='openstreetmap')

folium.Choropleth(
    geo_data = geojson,
    key_on = 'FIPS',
    name = 'choropleth',
    data = map_data,
    columns = ['FIPS','PctAllPop'],
    fill_color='OrRd',
    fill_opacity=0.8,
    line_opacity=0.2,
    nan_fill_color = 'grey',
    legend_name='Pct of population attending church'
    ).add_to(m1)

m1

I thought the problem might be that there are some FIPS values in the map_data table that are not in the geojson table, but I checked this with this code:

map_data[~map_data['FIPS'].isin(geojson['FIPS'])]

and it showed there were no unmatched FIPS values.

I thought the FIPS data were already string objects, but to double-check, I ran .astype() on each table:

map_data['FIPS'] = map_data['FIPS'].astype(str)

Even after doing this, folium is still reporting an error:

ValueError: key_on 'FIPS' not found in GeoJSON.

What am I am doing incorrectly?

Answer 1

So the problem may be as simple as I was using the wrong value for "key_on". This website says

The tricky part is the key_on parameter.

It should start with ‘feature’ and be in JavaScript objection notation, so in our case, it is feature.properties.ID_NIL

In my case, when I changed the key_on value to

key_on = 'feature.properties.FIPS'

I stopped getting the error. Frankly, I don't understand this, since my geo_data is in a GeoPandas table rather than a GeoJSON file (with dictionary-like structures, such as feature.properties).

Perhaps this is encoded behind the scenes in the GeoPandas table?