Bbox behavior different from spec

[bordax]

Current Behavior

When we initialize any GeoJson type without passing a bbox, it's setted to null.

from geojson_pydantic import FeatureCollection
from pprint import pprint

fc = {
    "type": "FeatureCollection",
    "features": [
        {
            "type": "Feature",
            "properties": "",
            "geometry": {
                "type": "Polygon",
                "coordinates": [
                    [
                        [-37.33175659179676, -9.424103736877383],
                        [-37.31844711303711, -9.420578002929688],
                        [-37.26972198486328, -9.45348930358881],
                        [-37.3108024597168, -9.467284202575684],
                        [-37.33392715454096, -9.487695693969727],
                        [-37.37298583984375, -9.524782180786133],
                        [-37.4452285766601, -9.476631164550724],
                        [-37.44975662231445, -9.453315734863281],
                        [-37.33175659179676, -9.424103736877383],
                    ]
                ],
            },
        }
    ],
}
fc = FeatureCollection(**fc)
pprint(fc.dict())

Will print

{'bbox': None,
 'features': [{'bbox': None,
               'geometry': {'bbox': None,
                            'coordinates': [[(-37.33175659179676,
                                              -9.424103736877383),
                                             (-37.31844711303711,
                                              -9.420578002929688),
                                             (-37.26972198486328,
                                              -9.45348930358881),
                                             (-37.3108024597168,
                                              -9.467284202575684),
                                             (-37.33392715454096,
                                              -9.487695693969727),
                                             (-37.37298583984375,
                                              -9.524782180786133),
                                             (-37.4452285766601,
                                              -9.476631164550724),
                                             (-37.44975662231445,
                                              -9.453315734863281),
                                             (-37.33175659179676,
                                              -9.424103736877383)]],
                            'type': 'Polygon'},
               'id': None,
               'properties': {},
               'type': 'Feature'}],
 'type': 'FeatureCollection'}

Expected Behavior

From the standard, at the definition of a GeoJSON Object:

GeoJSON Object

A GeoJSON object represents a Geometry, Feature, or collection of Features.
A GeoJSON object is a JSON object.
A GeoJSON object has a member with the name "type". The value of the member MUST be one of the GeoJSON types.
A GeoJSON object MAY have a "bbox" member, the value of which MUST be a bounding box array (see Section 5).

Then, from the section on bounding boxes:

5. Bounding Box

   A GeoJSON object MAY have a member named "bbox" to include
   information on the coordinate range for its Geometries, Features, or
   FeatureCollections. The value of the bbox member MUST be an array of
   length 2*n where n is the number of dimensions represented in the
   contained geometries, with all axes of the most southwesterly point
   followed by all axes of the more northeasterly point. The axes order
   of a bbox follows the axes order of geometries.

In other words, when passing no bbox, the parsed geometry should contain no bbox fields.

{
 'features': [{
               'geometry': {
                            'coordinates': [[(-37.33175659179676,
                                              -9.424103736877383),
                                             (-37.31844711303711,
                                              -9.420578002929688),
                                             (-37.26972198486328,
                                              -9.45348930358881),
                                             (-37.3108024597168,
                                              -9.467284202575684),
                                             (-37.33392715454096,
                                              -9.487695693969727),
                                             (-37.37298583984375,
                                              -9.524782180786133),
                                             (-37.4452285766601,
                                              -9.476631164550724),
                                             (-37.44975662231445,
                                              -9.453315734863281),
                                             (-37.33175659179676,
                                              -9.424103736877383)]],
                            'type': 'Polygon'},
               'id': None,
               'properties': {},
               'type': 'Feature'}],
 'type': 'FeatureCollection'
}

[vincentsarago]

For .dict() I don't think the behaviour is wrong, but it might for .json()

fc.json()
'{"type": "FeatureCollection", "features": [{"type": "Feature", "geometry": {"type": "Polygon", "coordinates": [[[-37.33175659179676, -9.424103736877383], [-37.31844711303711, -9.420578002929688], [-37.26972198486328, -9.45348930358881], [-37.3108024597168, -9.467284202575684], [-37.33392715454096, -9.487695693969727], [-37.37298583984375, -9.524782180786133], [-37.4452285766601, -9.476631164550724], [-37.44975662231445, -9.453315734863281], [-37.33175659179676, -9.424103736877383]]], "bbox": null}, "properties": {}, "id": null, "bbox": null}], "bbox": null}'

you can avoid this by using pydantic option exclude_none=True

fc.json(exclude_none=True)
'{"type": "FeatureCollection", "features": [{"type": "Feature", "geometry": {"type": "Polygon", "coordinates": [[[-37.33175659179676, -9.424103736877383], [-37.31844711303711, -9.420578002929688], [-37.26972198486328, -9.45348930358881], [-37.3108024597168, -9.467284202575684], [-37.33392715454096, -9.487695693969727], [-37.37298583984375, -9.524782180786133], [-37.4452285766601, -9.476631164550724], [-37.44975662231445, -9.453315734863281], [-37.33175659179676, -9.424103736877383]]]}, "properties": {}}]}

[bordax]

Yeah, it's ok that .dict() returns None values, but I think that the correct behavior on .json() shouldn't depends on an extra parameter. What do you think?

[vincentsarago]

Sadly I don't think there is an easy way to fix this on geojson-pydantic side 🤔

Note: this is also true for Feature id

[bordax]

Would be too aggressive to override .json() on _GeometryBase defaulting exclude_none or exclude_unset to True? hehe

[vincentsarago]

@eseglem do you think we should update the model_dump in geojson-pydantic 1.0 to take care of this?

[eseglem]

@vincentsarago Probably worth taking a look. Optional keywords that are not actually nullable are still not great in Pydantic.

I don't think I like messing with exclude_none since there are nullable required fields but exclude_unset might be reasonable. Not really a fan of messing with the default Pydantic behavior, since it would also affect things like Properties.

I think the best bet might be @model_serializer which we should be able to go in and say when serializing out remove the specific keywords that shouldn't be null.

[SamuelWillis]

Are there any plans to make these changes to 0.6.3 as well? Unfortunately I am unable to bump to pydantic 2 at this time and would like to avoid the null bbox values as well.

[vincentsarago]

@SamuelWillis we don't have any plan to backport this feature to 0.6.x but I'm happy to review any PR 🙏

[SamuelWillis]

@vincentsarago I'll see if I can find time here, I'm fairly new to python but I assume the solutions will be similar/the same?