Remove empty geometries from geodataframe

Question

Since geopandas can't perform overlay/intersection with lines, I found a workaround:

import geopandas as gpd
import fiona
from shapely.geometry import shape, MultiPolygon, LineString, MultiLineString


lines = gpd.read_file('lines.shp')
polyons= gpd.read_file('polygons.shp')
pol_f = fiona.open('polygons.shp')

pol_shp = MultiPolygon([shape(x["geometry"]) for x in pol_f])

lines2 = lines.geometry.apply(lambda x: x.intersection(pol_shp))
lines2.to_file('lines_int.shp')

But it does not let me because there are the empty geometries are GeometryCollections.

I tried this to filter, but it won't work.

lines3 = lines2[lines2.geometry != np.nan]
lines3 = lines2[lines2.geometry == LineString]
lines3 = lines2[lines2.geometry == LineString()]
lines3 = lines2[lines2.geometry == MultiLineString()]
lines3 = lines2[lines2.geometry == MultiLineString]
lines3 = lines2[lines2.geometry != GeometryCollection]

Any ideas?

Answer 1

The GeoDataFrame and GeoSeries have an is_empty attribute, similar to shapely geometry objects. You can use this to filter out the empty ones:

In [7]: lines = geopandas.GeoSeries([
   ...:     LineString([(0, 0), (1, 1)]), LineString([]), LineString([(1, 1), (2, 2)])])

In [8]: lines                             
Out[8]: 
0       LINESTRING (0 0, 1 1)
1    GEOMETRYCOLLECTION EMPTY
2       LINESTRING (1 1, 2 2)
dtype: object

In [9]: lines.is_empty                                      
Out[9]: 
0    False
1     True
2    False
dtype: bool

In [10]: lines[~lines.is_empty]                                            
Out[10]: 
0    LINESTRING (0 0, 1 1)
2    LINESTRING (1 1, 2 2)
dtype: object

Answer 2

I'm commenting this post 3 years after @joris answer. In GeoPandas v0.6.0, the missing data handling was refactored and made more consistent across the library. From this version on, GeoSeries.isna() no longer treating empty as missing:

#For GeoPandas<v0.6.0
print(s)
0    POLYGON ((0 0, 1 1, 0 1, 0 0))
1                              None
2          GEOMETRYCOLLECTION EMPTY
dtype: object

print(s.isna())
0    False
1     True
2     True
dtype: bool

#for GeoPandas >=v0.6.0
s.isna()
0    False
1     True
2    False
dtype: bool

As today, I looked into the documentation and I did not find a specific method to remove empty geometries, if not "improvising" a solution. If anyone wants to contribute to this topic, please share.