Giving variables and functions good names is important. But even an awesome name for a dataframe isn't very useful. So I created an annotation that allows "documenting" dataframes that functions get as input or produce as output. You can tell which columns to expect or even use Pydantic models to validate each row.