Currently spans is the only way to get the filename, which is useful given that there will be an API to load and parse only specific files.