I frequently find myself using the following pattern when collecting outputs of a dynamic fan

@op
def join_data_and_add_calculated_columns(dataframes:List[CustomDataFrameType1])->CustomDataFrameType2:
    if not dataframes:
        return pd.DataFrame()
    full_df = pd.concat(dataframes)

Before I started using custom types created by

create_dagster_pandas_dataframe_type

this was no problem. Now its an issues because

return pd.DataFrame()

does not pass validation. What's the recommended behaviour when you don't have inputs and want to return an empty dataframe?

The type Dagster.Type.Nothing can help, I think

Do you rely on the fact that it is an empty dataframe in your downstream ops? I wonder if you might be able to conditionally return

None

by making the datatype Optional