Description of field of schema model does not make it into FastAPI openapi.json

[BigNerd]

When defining a schema model as follows I would love to see the description occur in the FastAPI openapi.json / Swagger UI:

class InputSchema(pa.SchemaModel):
    year: Series[int] = pa.Field(description="the year pandera was released")

Is this already possible by any means?

[cosmicBboy]

hi @BigNerd ! we'll need to update this function so that the json schema representation contains the description and other metadata like title, etc. https://github.com/unionai-oss/pandera/blob/master/pandera/json_schema.py

Would you be open to making a contribution on this front?

[BigNerd]

Hi @cosmicBboy yes, I think I can give it a try during the next weeks or so. Thanks for pointing me to the piece of code that needs to be worked on for this.

[BigNerd]

Here is the change to to_json_schema() that I want to contribute when I have time to go through the fork, branch, pull request process.

def to_json_schema(dataframe_schema):
    """Serialize schema metadata into json-schema format.

    :param dataframe_schema: schema to write to json-schema format.

    .. note::

        This function currently does not fully specify a pandera schema,
        and is primarily used internally to render OpenAPI docs via the
        FastAPI integration.
    """
    empty = pd.DataFrame(columns=dataframe_schema.columns.keys()).astype(
        {k: v.type for k, v in dataframe_schema.dtypes.items()}
    )
    table_schema = pd.io.json.build_table_schema(empty)

    def _field_json_schema(field):
        annotations = {}
        name = field["name"]
        columns = dataframe_schema.columns
        if name in columns and columns[name].title is not None:
            annotations["title"] = columns[name].title
        if name in columns and columns[name].description is not None:
            annotations["description"] = columns[name].description
        return {
            "type": "array",
            "items": {
                "type": field["type"],
                **annotations,
            },
        }

    return {
        "title": dataframe_schema.name or "pandera.DataFrameSchema",
        "type": "object",
        "properties": {
            field["name"]: _field_json_schema(field)
            for field in table_schema["fields"]
        },
    }

[cosmicBboy]

cool! I can review if you can make a PR ⭐️