Support orient keyword in to_json

[paddymul]

• [X ] I have checked that this issue has not already been reported.

• [X ] I have confirmed this bug exists on the latest version of geopandas.

• (optional) I have confirmed this bug exists on the main branch of geopandas.

---

Note: Please read this guide detailing how to provide the necessary information for us to reproduce your bug.

Code Sample, a copy-pastable example

# Your code here
from shapely.geometry import Point
import geopandas
d = {'col1': ['name1', 'name2'], 'geometry': [Point(1, 2), Point(2, 1)]}
gdf = geopandas.GeoDataFrame(d, crs="EPSG:3857")
gdf.to_json(orient='table')

Problem description

I have a table library that supports pandas. I expected to be able to serialize a GeoDataFrame with the same serializer methods, but to_json doesn't support orient at all.

I understand adding extra arguments for geo specific features, but I would expect pandas default arguments to work.

Expected Output

given that pandas of

import pandas as pd
print(pd.DataFrame({'a': [{'foo':9, 'bar':10}, {'c':9}]}).to_json(orient='table', indent=2))

outputs as

{
  "schema":{
    "fields":[  { "name":"index",  "type":"integer"  },   {  "name":"a",    "type":"string"  } ],
    "primaryKey":[ "index"  ],
    "pandas_version":"1.4.0"
  },
  "data":[
    { "index":0,  "a":{ "foo":9, "bar":10  }  },
    { "index":1,  "a":{ "c":9  }
    }
  ]
}

I would expect the following GeoPandas to output as

d = {'col1': ['name1', 'name2'], 'geometry': [Point(1, 2), Point(2, 1)]}
gdf = geopandas.GeoDataFrame(d, crs="EPSG:3857")
gdf.to_json(orient='table')

to output as

{
    "type": "FeatureCollection",
    "schema":{
        "fields":[  { "name":"id",       "type":"integer"  },
                    { "name":"col1",     "type":"string"  },
                    { "name":"geometry", "type": "FeatureCollection"}],
        "primaryKey":[ "id"  ],
        "pandas_version":"geo-1.4.0"
  },

    "data": [
        {"col1": "name1",
         "id": "0",
         "features": {"type": "Feature",
                      "geometry": {
                          "type": "Point",
                          "coordinates": [1.0, 2.0 ]}}},
        {"col1": "name2",
         "id": "1",
         "features": {"type": "Feature",
                      "geometry": {
                          "type": "Point",
                          "coordinates": [2.0, 1.0 ]}}}
    ]
}
 

Output of geopandas.show_versions()

SYSTEM INFO

python : 3.10.12 (main, Nov 20 2023, 15:14:05) [GCC 11.4.0]
executable : /usr/bin/python3
machine : Linux-6.1.58+-x86_64-with-glibc2.35

GEOS, GDAL, PROJ INFO

GEOS : 3.11.2
GEOS lib : None
GDAL : 3.6.4
GDAL data dir: /usr/local/lib/python3.10/dist-packages/fiona/gdal_data
PROJ : 9.3.0
PROJ data dir: /usr/local/lib/python3.10/dist-packages/pyproj/proj_dir/share/proj

PYTHON DEPENDENCIES

geopandas : 0.13.2
numpy : 1.25.2
pandas : 1.5.3
pyproj : 3.6.1
shapely : 2.0.2
fiona : 1.9.5
geoalchemy2: None
geopy : 2.3.0
matplotlib : 3.7.1
mapclassify: None
pygeos : None
pyogrio : None
psycopg2 : 2.9.9 (dt dec pq3 ext lo64)
pyarrow : 10.0.1
rtree : None

[m-richards]

Thanks @paddymul for opening this issue. I wouldn't say this is a bug per se, but a mismatch of expectations. In the geospatial world the standard json representation of features is GeoJson - which specifies a specific layout for storing features separate to geometry alongside storing spatial metadata like a CRS (coordinate reference system).

(For instance compare the output of gdf.to_json(indent=2) and pd.DataFrame(gdf.to_wkt()).to_json(indent=2) in your example).

In principle I think we would be open to supporting other orientations, so long as that does not lead to confusion about what is a generic interoperable geojson format and what is a pandas specific format. It'd be good for other maintainers to weight in, but my initial proposal to support this would be to introduce an orient="GeoJSON" which is the default value and support the other values via pandas. Geometry would be exported as WKT and we'd probably need to emit a warning about the loss of CRS metadata (and then perhaps a second stage would be to agree on a way to embed CRSs if there was an appetite for that.

[paddymul]

FWIW I was able to get a regular pandas dataframe with the following code:

pd_df = pd.DataFrame(dict(zip(df.columns, df.to_numpy().T)))

Then I serialize via existing pandas methods.

I no longer need geopandas to change the behavior of to_json, but I figure this code could be useful for others who stumble across this problem.

[m-richards]

FWIW I was able to get a regular pandas dataframe with the following code:

pd_df = pd.DataFrame(dict(zip(df.columns, df.to_numpy().T)))

I haven't compared the performance, but I think a slightly clearer way to express this would be pd.DataFrame(gdf).astype({"geometry":"object"}) (or a generalisation of it if there are multiple geometry columns), but it is more verbose.