Add to_csv() method to GeoDataFrame for point geometries about geopandas HOT 12 OPEN

@bidhya Ah, I see what you mean. The current to_csv function in geopandas, including my branch above, is just the pandas function. It hasn't been overridden with a GeoDataFrame-specific method. So going in a to_csv/from_csv cycle will result in the behavior you describe. I think a GeoDataFrame.to_csv method would be good though. Had there been more activity in this thread after my first comment, six months ago, I would have attempted such a method and submitted a PR.

from geopandas.

I'd also like to see a from_csv method in this same vein. The only potential 'issue' is that the Pandas API allows for returning a text streamer, which will require a bit more thinking/wrapping.

from geopandas.

Actually, OGR does support CSV files with WKT geometries... so you could output geometries other than points...

from geopandas.

Ah. I've never seen non-point CSVs in the wild but I have a pretty limited view.

Maybe make arguments x and y optional? If they're specified, then it would behave as above and assume points. Otherwise output WKT...?

from geopandas.

I think it'd be great to optionally output VRT files alongside the CSVs. The VRT is a simple XML document which can be read by OGR to infer the geographic data contained in a CSV, and makes interoperating with CSVs much easier. See this GIS.SE question for a look at the approach, or the CSV driver docs.

from geopandas.

I am also interested in to_csv/from_csv methods. I wrote a little from_csv function (which wraps a from_dataframe function) in my fork (ddefbd4). It makes no attempt to handle polygons, and requires that the column names for x/y be specified. It works fine for my purposes, but I have not tested rigorously.

(Edited SHA after switching commit to a new branch; I forgot that commits get automatically included in existing pull requests.)

from geopandas.

So it is possible to use from_csv to load the saved geodataframe without any hack?

from geopandas.

I'm not quite sure what you mean by "without any hack". Assuming a csv file with points whose columns are labeled longitude and latitude, it would create a GeoDataFrame with the following single call:

>>> gdf = geopandas.GeoDataFrame.from_csv("filename.csv", "longitude", "latitude")

However, I never made a pull request, since this issue was already here, so my implementation is not currently part of geopandas.

from geopandas.

@jdmcbr In geopandas, the geometry information is embedded in the column names "geomtry". When I save this geodataframe, this geometry column is saved as well (there is no explicit logitude and latitude columns).

So when I load this csv file, I expect it to work exactly like the original geodataframe that saved this. But this is not how it works. For example, I load the saved dataframe as follows

gdf = geopandas.GeoDataFrame.from_csv("filename.csv")
Now using a plot command will plot the numerical columns like regular pandas dataframe. If gdf we true geodataframe, it should have plotted the geometry only.

Thats what I mean by "without a hack". When you load a shapefile with geopandas.read_file('filename.shp') there is no column called longitude and latitude, only a column called geometry. As soon as I save this geodataframe to csv and load again, it stops behaving as a geodataframe.

from geopandas.

@jwass @jdmcbr May I ask if this problem was resolved later？I also encountered a similar problem.

Expected behavior and actual behavior.

File "/home/work/extoanaservice/app/analyzers/work_area.py", line 627, in init
mesh_link_gdf.to_csv(os.path.join(self.file_manager.storage_folder(), mesh_id + '.csv'), index=False)
File "/home/work/extoanaservice/Python-2.7.14-env/lib/python2.7/site-packages/pandas/core/frame.py", line 1745, in to_csv
formatter.save()
File "/home/work/extoanaservice/Python-2.7.14-env/lib/python2.7/site-packages/pandas/io/formats/csvs.py", line 151, in save
self._save()
File "/home/work/extoanaservice/Python-2.7.14-env/lib/python2.7/site-packages/pandas/io/formats/csvs.py", line 263, in _save
self._save_chunk(start_i, end_i)
File "/home/work/extoanaservice/Python-2.7.14-env/lib/python2.7/site-packages/pandas/io/formats/csvs.py", line 277, in _save_chunk
quoting=self.quoting)
File "/home/work/extoanaservice/Python-2.7.14-env/lib/python2.7/site-packages/pandas/core/internals.py", line 766, in to_native_types
values = values.astype(str)
File "/home/work/extoanaservice/Python-2.7.14-env/lib/python2.7/site-packages/shapely/geometry/base.py", line 238, in str
return self.wkt
File "/home/work/extoanaservice/Python-2.7.14-env/lib/python2.7/site-packages/shapely/geometry/base.py", line 367, in wkt
return WKTWriter(lgeos, **kw).write(self)
File "/home/work/extoanaservice/Python-2.7.14-env/lib/python2.7/site-packages/shapely/geos.py", line 337, in init
self._writer = self._lgeos.GEOSWKTWriter_create()
ArgumentError: argument 1: <class 'ctypes.ArgumentError'>: argument 1: <class 'ctypes.ArgumentError'>: ......

There is a certain probability that this problem will occur.

Shapely version and provenance

Shapely==1.6.4.post1 fiona==1.8.22 pandas==0.23.0 geopandas==0.6.0
libgeos 3.6.2

from geopandas.

to_csv has not been implemented. The recommended, geometry agnostic way, is to cast geometries to WKT before saving as csv.

gdf.to_wkt().to_csv()

from geopandas.

@githubtianfeng Please see the answers in the shapely repository regarding your environment issues. None is f the versions you have is supported any longer.

from geopandas.