GeoPandas setting geometry for a row with .loc and Multipolygons

Question

Given the following example:

def merge_with_largest(input_gdf,poly):
    
    # select the polygons bordering polygon
    close_by_gdf = input_gdf[input_gdf.geometry.intersects(poly)]
    # selects the index of the row with the largest area
    selected_index = close_by_gdf['area'].argmax()
    # now we merge the geometry of the selected index with the polygon geometry
    selected_geom = input_gdf.loc[selected_index]['geometry']
    # setting the 
    input_gdf.loc[selected_index,['geometry']] = gpd.GeoSeries((selected_geom,poly)).unary_union

What I want to do here is to merge the geometry of poly with the largest of the polygons that do intersect it with. I want to do this by using .unary_union between the two selected polygons and then update the geometry of the selected polygon with what .unary_union returns. The problem though is that sometimes .unary_union returns a mulitpolygon and then the .loc method throws an error because it's an iterable.

ValueError: Must have equal len keys and value when setting with an iterable

Is there a way to overwrite this and set the multipolygon as the the geometry for the selected column?

Answer 1

I found a workaround to my problem credit here goes to github https://github.com/geopandas/geopandas/issues/992

the solution in a nutshell:

geom = df.loc[1, 'geometry']
df.loc[[0], 'geometry'] = geopandas.GeoSeries([geom])

if geom is a multipolygon we cast it into a GeoSeries and then set the series as the new geometry

Answer 2

I'm proposing another solution that solved a similar problem for me.

In my case, I had some Features in a GeoDataFrame with shared attributes and adjacent geometries, which could be consolidated as one Feature with a union.

The general solution is to create a temporary GeoDataFrame and using the GeoDataFrame.update() method. Important to note that the temporary GeoDataFrame needs to have a matching properties so it can be joined to update the original Feature.

Code:

original_gdf = gpd.read_file('geojson_filepath.geojson')

shared_id_attribute = 'unique_to_two_features'

idx_1, idx_2 = original_gdf.loc[original_gdf['aoid_id'].str.contains(shared_id_attribute), :].index
    
geom_1 = original_gdf.loc[idx_1, 'geometry']
geom_2 = original_gdf.loc[idx_2, 'geometry']
    
union_geom = geom_1.union(geom_2)
    
temp_df = gpd.GeoDataFrame({'aoi_id': aoi_id,
                            'geometry': [union]}, # need to wrap a list here
                           geometry='geometry', 
                           crs=original_gdf.crs)

# set an index to the unique identifier you want to join on
# otherwise `GeoDataFrame.update()` will update rows with matching index and column values. 
temp_df.index = temp_df['aoi_id']
    
# update geom 1 with union of both
original_gdf.update(temp_df)

assert original_gdf.loc[idx_1, 'geometry'] != None

original_gdf.drop(labels=idx_2, axis=0, inplace=True)