Allow serialization overwrite, or different serialization on specific component/props

[LeoGrosjean]

Describe the bug
Can't overwrite an existing serializer for a defined type (here pd.DataFrame

To Reproduce

@rx.serializer
def serialize_dataframe(dataframe: pd.DataFrame) -> List[Dict[str, Any]]:
        """
        Serialize a pandas DataFrame to dict with records
        """
        return dataframe.to_dict("records")

usage
In my vega/altair wrapper I need to serialize dataframe to records (list of dict[col, value]

quickfix fail

def serializer(
    fn: Serializer | None = None,
    to: Type | None = None,
    overwrite: bool = False, # <- yuhu
) -> Serializer:
        ...
        if registered_fn is not None and (registered_fn != fn and not overwrite):
        raise ValueError(
            f"Serializer for type {type_} is already registered as {registered_fn.__qualname__}."
        )

But when app is starting, serializer is called with overwrite default value ! So False instaid of True !

[adhami3310]

Why is not possible to use the value of the current serializer of dataframe?

[LeoGrosjean]

I wrapped altair/vega component here

But vega-lite/vega needs data like this :

df.to_dict("records")
[
  {'date': Timestamp('2012-01-01 00:00:00'),
  'precipitation': 0.0,
  'temp_max': 12.8,
  'temp_min': 5.0,
  'weather': 'drizzle',
  'wind': 4.7},
  {'date': Timestamp('2012-01-02 00:00:00'),
  'precipitation': 10.9,
  'temp_max': 10.6,
  'temp_min': 2.8,
  'weather': 'rain',
  'wind': 4.5}
]

instead of

df.to_dict()
{'date': {0: Timestamp('2012-01-01 00:00:00'),
        1: Timestamp('2012-01-02 00:00:00')},
'precipitation': {0: 0.0, 1: 10.9},
'temp_max': {0: 12.8, 1: 10.6},
'temp_min': {0: 5.0, 1: 2.8},
'weather': {0: 'drizzle', 1: 'rain'},
'wind': {0: 4.7, 1: 4.5}}

Its not a major issue, just for convenience

[LeoGrosjean]

What do you think of something like :

DataFrameAltair= Annotated[pd.DataFrame, {"orient": "records "}]

And then in the serializer check for annotation ?

[LeoGrosjean]

I succeed !

Will tidy my code next week and add tests

Serializer is created, but not used :(

[adhami3310]

dang it, that seemed like a good solution

is it possible to wrap dataframe some other way? like with an actual type and not an alias