In read_json we parse columns as dates if they have column names that look like they might be dates e.g. created_at, date, something_date. It's a cheeky solution ... Perhaps we could pop this same logic into read_csv?

Comment From: TomAugspurger

Too magical I think.