Creating Weights from dataframe: IndexError: index 0 is out of bounds for axis 0 with size 0

[raholler]

I get the following error when running

W = libpysal.weights.KNN.from_dataframe(df,  k=4)

File Anaconda3\envs\geo_new\lib\site-packages\libpysal\weights\distance.py:905, in DistanceBand.from_dataframe(cls, df, threshold, geom_col, ids, **kwargs)
    903 if geom_col is None:
    904     geom_col = df.geometry.name
--> 905 pts = get_points_array(df[geom_col])
    906 if ids is None:
    907     ids = df.index.tolist()

File Anaconda3\envs\geo_new\lib\site-packages\libpysal\weights\util.py:1081, in get_points_array(iterable)
   1078 try:
   1079     if HAS_SHAPELY:
   1080         data = np.vstack(
-> 1081             [
   1082                 np.array(shape.centroid.coords)[0]
   1083                 if isinstance(shape, BaseGeometry)
   1084                 else np.array(shape.centroid)
   1085                 for shape in first_choice
   1086             ]
...
   1092                 for shape in first_choice
   1093             ]
   1094         )

IndexError: index 0 is out of bounds for axis 0 with size

The error disappears, when I restrict the geodataframe to areas without islands.

• Platform information: nt win32
• Python version: 3.10.6
• SciPy version: 1.9.1
• NumPy version: 1.23.2
• Geopandas: 0.11.1
• libpysal: 4.6.2

test.zip

[sjsrey]

The issue seems to be that there are observations that do not have geometries:

Without coordinates, it isn't possible to calculate the distances needed for the knn determination.
If you drop those observations, all is well.

[sjsrey]

There are 11 such observations:

[raholler]

Thanks for the swift response. I am sorry, I actually checked in the originial data

data.query("geometry.isnull()")

and it seemed that there were no missing geometries. I will check again, what is going on.

[sjsrey]

Closing since the problem is not due to a bug.