feat(data frame): Add generic type support via render return value

[schloerke]

Add support for @render.data_frame to know what the real type should be for .data_value(), .data(), and any other data related methods.

Ex: Both first_df and second_df are aware that the return value of the render method has an underlying data type of [pandas.]DataFrame. This also works for polars.DataFrame.

# app.py
from palmerpenguins import load_penguins_raw  # pyright: ignore[reportMissingTypeStubs]

from shiny.express import render


@render.data_frame
def first_df():
    return render.DataGrid(data=load_penguins_raw())


@render.data_frame
def second_df():
    return first_df.data_view(selected=True)


@render.code
def code():
    return str(second_df.data())

TODO:

• Fix PandasCompatible type. The __new__ of DataGrid and DataTable does not convey that it is converted to a pandas.DataFrame.

  • Solution ✅ : Users call df.to_pandas() to convert their data to pandas
    

  • Bad workaround: Give type to @render.data_frame[PdDataFrame]