setting geometry with .loc from column value

• I have come across an issue that crops up when attempting to set geometry values with .loc using the values from another column. Using .loc to set geometry works normally when setting with the index value, but throws an error (see the MWE/traceback below) otherwise. I don’t think this is expected behavior, but I could be very wrong.

• Being able to set geometry from a non-index column value is critical for a specific workflow I am using, so for an interim solution I ginned up a super hacky method for teasing out the index value based on a column value to allow for using .loc.

• The error is actually being thrown from pandas, but only seems to be happening with shapely.geometry type objects so I wasn’t sure if geopandas was the correct place to post this, but I figured it would be a safe start.

• Version info:

platform - 'Darwin-17.7.0-x86_64-i386-64bit'
python - '3.6.6 | packaged by conda-forge | (default, Jul 26 2018, 09:55:02)
         '[GCC 4.2.1 Compatible Apple LLVM 6.1.0 (clang-602.0.53)]'
geopandas - '0.4.0'
pandas - '0.23.4'
shapely - '1.6.4.post1'

---

MWE

In [1]: gdf = gpd.GeoDataFrame()
In [2]: line_1 = LineString(((0,0), (1,1)))
In [3]: line_2 = LineString(((3,3), (5,5)))
In [4]: gdf['geometry'] = [line_1]
In [5]: gdf['line_id'] = '999'
In [6]: print(gdf)

                geometry line_id
0  LINESTRING (0 0, 1 1)     999

In [7]: print(gdf.loc[(gdf['line_id'] == '999'), 'geometry'])

0    LINESTRING (0 0, 1 1)
Name: geometry, dtype: object

[Fail] Attempt to set geometry value by the location of (gdf['line_id'] == '999')

In [8]: gdf.loc[(gdf['line_id'] == '999'), 'geometry'] = line_2

ValueError                                Traceback (most recent call last)
<ipython-input-154-8d7c73a5b6c8> in <module>()
----> 1 gdf.loc[(gdf['line_id'] == '999'), ['geometry']] = line2

~/miniconda3/envs/py3_at1866/lib/python3.6/site-packages/pandas/core/indexing.py in __setitem__(self, key, value)
    187             key = com._apply_if_callable(key, self.obj)
    188         indexer = self._get_setitem_indexer(key)
--> 189         self._setitem_with_indexer(indexer, value)
    190 
    191     def _validate_key(self, key, axis):

~/miniconda3/envs/py3_at1866/lib/python3.6/site-packages/pandas/core/indexing.py in _setitem_with_indexer(self, indexer, value)
    649             self.obj._consolidate_inplace()
    650             self.obj._data = self.obj._data.setitem(indexer=indexer,
--> 651                                                     value=value)
    652             self.obj._maybe_update_cacher(clear=True)
    653 

~/miniconda3/envs/py3_at1866/lib/python3.6/site-packages/pandas/core/internals.py in setitem(self, **kwargs)
   3691 
   3692     def setitem(self, **kwargs):
-> 3693         return self.apply('setitem', **kwargs)
   3694 
   3695     def putmask(self, **kwargs):

~/miniconda3/envs/py3_at1866/lib/python3.6/site-packages/pandas/core/internals.py in apply(self, f, axes, filter, do_integrity_check, consolidate, **kwargs)
   3579 
   3580             kwargs['mgr'] = self
-> 3581             applied = getattr(b, f)(**kwargs)
   3582             result_blocks = _extend_blocks(applied, result_blocks)
   3583 

~/miniconda3/envs/py3_at1866/lib/python3.6/site-packages/pandas/core/internals.py in setitem(self, indexer, value, mgr)
    938         # set
    939         else:
--> 940             values[indexer] = value
    941 
    942         # coerce and try to infer the dtypes of the result

ValueError: shape mismatch: value array of shape (2,2) could not be broadcast to indexing result of shape (1,1)

[Pass] Attempt to set geometry value by index location

In [9]: gdf.loc[0, 'geometry'] = line_2
In [10]: print(gdf)

                geometry line_id
0  LINESTRING (3 3, 5 5)     999

[martinfleis]

I would say so, yes. I think it somehow related to the fact that np.array(LineString(((0,0), (1,1), (1,3), (2,3), (4,5)))) returns array of coordinates, but that is just a guess. Thanks!

[jGaboardi]

migrated to pandas-dev/pandas#28924