Do not want to include None attributes to JSON.

[daigok]

Is your feature request related to a problem? Please describe.
Not related to a problem.

Describe the solution you'd like
I do not want to include None attributes to JSON result. FastAPI has an include_none option on jsonable_encoder method. However, there is no way to set this option False.

Describe alternatives you've considered
Manually uses jsonable_encoder with include_none=True.

Additional context
If FastAPI provides a way to globally configure json encoder. It would be handy.

[dmontagu]

For what it’s worth, I believe pydantic 1.0 is aiming to support custom json encoders per model. That may go a long way toward powering this feature (though may still require some changes to fastapi).

[tiangolo]

@daigok you can also use:

response_model_include=None
response_model_exclude=set()
response_model_skip_defaults=None
response_model_exclude_unset

in the path operation decorator, like:

@app.get("/", response_model_skip_defaults=None)
def read():
    ...

[vincentsarago]

👋 I'm not sure if it's the best place but I'm looking at a similar thing.

I've got a model with some default values, when adding response_model_exclude_unset=True those default values get skiped and the output has only the value I passed while I wanted to get a mix of default + input.

class TileJSON(BaseModel):
    """
    TileJSON model.

    Based on https://github.com/mapbox/tilejson-spec/tree/master/2.2.0

    """

    tilejson: str = "2.2.0"
    name: Optional[str]
    description: Optional[str]
    version: str = "1.0.0"
    attribution: Optional[str]
    template: Optional[str]
    legend: Optional[str]
    scheme: str = "xyz"
    tiles: List[str]
    grids: Optional[List[str]]
    data: Optional[List[str]]
    minzoom: int = Field(0, ge=0, le=30)
    maxzoom: int = Field(30, ge=0, le=30)
    bounds: List[float] = [-180, -90, 180, 90]
    center: Tuple[float, float, int]

Thanks for pointing me to the right doc of code, and I'm happy to contribute in anyway

[dmontagu]

@vincentsarago can you explain what you are receiving as input, what you want to get as output, and what you are actually getting as output?

[vincentsarago]

sure @dmontagu
so this is based on tileJSON specification. tilejson=2.2.0 is required and thus I set it with default values in the model. minzoom. maxzoom, bounds are optional but with default values so I expected them to be in my output body.

Based in the model shared before, here is what I'm doing

@router.get(
    "/tilejson.json",
    response_model=TileJSON,
    responses={200: {"description": "Return a tilejson"}},
)
def tilejson():
    # constructing the tile url here 
    tile_url = "http://127.0.0.1:8000/v1/{z}/{x}/{y}"
    # Set some variable
    bounds = [-70.36082319774525, 47.75740353126241, -68.86721739002996, 48.7530129648102]
    minzoom = 1
    maxzoom = 5
    
    return dict(
        tiles=[tile_url],
        bounds=bounds,
        minzoom=minzoom,
        maxzoom=maxzoom,
    )

And here is the output

{
  "tiles": ["http://127.0.0.1:8000/v1/{z}/{x}/{y}"],
  "minzoom": 1,
  "maxzoom": 5,
  "bounds": [
    -70.36082319774525, 47.75740353126241, -68.86721739002996, 48.7530129648102
  ],
}

while idealy it should be (with defaults values from the model)

{
  "tilejson": "2.2.0",
  "tiles": ["http://127.0.0.1:8000/v1/{z}/{x}/{y}"],
  "minzoom": 1,
  "maxzoom": 5,
  "scheme": "xyz",
  "bounds": [
    -70.36082319774525, 47.75740353126241, -68.86721739002996, 48.7530129648102
  ],
}

I understand that I could use response_model_include={"tilejson", "scheme" ...} to get this but I was surprise that default values wheren't forwarded.

[dmontagu]

I see, that's helpful. I think longer term exclude_unset is the way to go here, but will require a little bit of a rework of how serialization happens.

In the short term I'm afraid manual handling is probably going to be necessary for this case.

One thing that might be convenient, depending on context, would be to override the dict or _iter or similar methods to remove keys with value None, if that is your goal. You could do that in a base class and get the logic reused. I believe that would result in FastAPI serializing the model the way you want. Obviously not ideal longer term (and maybe not even workable in the short term depending on how you are using things elsewhere), but might be viable for now?

[vincentsarago]

👼 thanks for your answer @dmontagu, really appreciate it. I'll see what I can do on my end and try to share the method.

[tiangolo]

Thanks for the help @dmontagu !

@daigok is your problem solved? May we close this issue?

[daigok]

@tiangolo
Thank you. Please close the issue.

[github-actions[bot]]

Assuming the original need was handled, this will be automatically closed now. But feel free to add more comments or create new issues or PRs.