There are a couple things we can do here: - Add `_file` of `_reader` methods that accept a file name or a reader rather than a loaded string, this would save a large copy - Do the parsing in parallel