Cannot write csv in 0.6.0 if geometry contains multipolygon

I have encountered an error in 0.6.0 when writing a GeoDataFrame to csv if the geometry has a MultiPolygon shape.

This worked in 0.5.1:

$ conda create -n gpd051 python=3.6 geopandas=0.5.1
<conda output>
$ conda activate gpd051
(gpd051) $ python
Python 3.6.7 | packaged by conda-forge | (default, Jul  2 2019, 02:07:37)
[GCC 4.2.1 Compatible Clang 4.0.1 (tags/RELEASE_401/final)] on darwin
Type "help", "copyright", "credits" or "license" for more information.
>>> from shapely.wkt import loads
>>> import geopandas as gpd
>>> geom_str = 'MULTIPOLYGON (((1 0, 1 1, 0 1, 0 0, 1 0)), ((3 0, 3 1, 2 1, 2 0, 3 0)))'
>>> geom = loads(geom_str)
>>> gdf = gpd.GeoDataFrame(geometry=[geom])
>>> gdf.to_csv()
',geometry\n0,"MULTIPOLYGON (((1 0, 1 1, 0 1, 0 0, 1 0)), ((3 0, 3 1, 2 1, 2 0, 3 0)))"\n'

But the same steps fail in 0.6.0 with ValueError: setting an array element with a sequence:

$ conda create -n gpd060 python=3.6 geopandas=0.6.0
<conda output>
$ conda activate gpd060
(gpd060) $ python
Python 3.6.7 | packaged by conda-forge | (default, Jul  2 2019, 02:07:37)
[GCC 4.2.1 Compatible Clang 4.0.1 (tags/RELEASE_401/final)] on darwin
Type "help", "copyright", "credits" or "license" for more information.
>>> from shapely.wkt import loads
>>> import geopandas as gpd
>>> geom_str = 'MULTIPOLYGON (((1 0, 1 1, 0 1, 0 0, 1 0)), ((3 0, 3 1, 2 1, 2 0, 3 0)))'
>>> geom = loads(geom_str)
>>> gdf = gpd.GeoDataFrame(geometry=[geom])
>>> gdf.to_csv()
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "/Users/john.flavin/anaconda3/envs/gpd060/lib/python3.6/site-packages/pandas/core/generic.py", line 3228, in to_csv
    formatter.save()
  File "/Users/john.flavin/anaconda3/envs/gpd060/lib/python3.6/site-packages/pandas/io/formats/csvs.py", line 202, in save
    self._save()
  File "/Users/john.flavin/anaconda3/envs/gpd060/lib/python3.6/site-packages/pandas/io/formats/csvs.py", line 324, in _save
    self._save_chunk(start_i, end_i)
  File "/Users/john.flavin/anaconda3/envs/gpd060/lib/python3.6/site-packages/pandas/io/formats/csvs.py", line 340, in _save_chunk
    quoting=self.quoting,
  File "/Users/john.flavin/anaconda3/envs/gpd060/lib/python3.6/site-packages/pandas/core/internals/blocks.py", line 760, in to_native_types
    values = values.astype(str)
ValueError: setting an array element with a sequence

I have a hypothesis about what’s going on. (I don’t know much about the internals of how GeoDataFrames are represented, so take this with a grain of salt.) The source of the exception is this block in pandas/core/internals/blocks.py:

        if not self.is_object and not quoting:
            values = values.astype(str)
        else:
            values = np.array(values, dtype="object")

My guess is that since the geometry column used to be an object type, not self.is_object would have been False and values = values.astype(str) would not have been evaluated. But now geometry is of type geometry so not self.is_object is True and we get the error.

[jorisvandenbossche]

@johnflavin I opened a PR to fix the issue on the pandas side (https://github.com/pandas-dev/pandas/pull/28841), and a PR in geopandas to fix the astype issue (https://github.com/geopandas/geopandas/pull/1147).

This means that for your exact example to work, you will need to wait until the next pandas release (or stay on geopandas 0.5). But with the fix in geopandas (to be released in 0.6.1), you will already be able to work-around it with first casting your geometry column to strings before calling to_csv:

gdf.astype({'geometry': str}).to_csv()

[jorisvandenbossche]

@johnflavin FYI 0.6.1 is released