These functions are currently un-annotated, so some type checkers (e.g. pyright) will try to infer based on the underlying function calls like load_data and by extension function calls like pd.DataFrame.
It would be better if we explicitly annotated these with typing.Any since it would mean the type checker short-circuits this inference and wouldn't raise an error if the user fails to narrow down the type through case-checking from DataFrame | dict etc.
It would also play nicer with extension packages like geopins which extend the interface to give extra return types dynamically, which the type checker doesn't know about.
I understand that pins isn't typically used in the context of heavy-duty type checking but I think this is pretty easy thing to add and would make it more pleasant to work with in VS Code which usually does light pyright analysis via pylance.
I can put in a PR
These functions are currently un-annotated, so some type checkers (e.g. pyright) will try to infer based on the underlying function calls like
load_dataand by extension function calls likepd.DataFrame.It would be better if we explicitly annotated these with
typing.Anysince it would mean the type checker short-circuits this inference and wouldn't raise an error if the user fails to narrow down the type through case-checking fromDataFrame | dictetc.It would also play nicer with extension packages like
geopinswhich extend the interface to give extra return types dynamically, which the type checker doesn't know about.I understand that pins isn't typically used in the context of heavy-duty type checking but I think this is pretty easy thing to add and would make it more pleasant to work with in VS Code which usually does light pyright analysis via pylance.
I can put in a PR