using `with` to be consistent with other code in datasets might also be useful in case file is not present due to corruption of something else
