It'd be awesome to add the equivalent to error_bad_lines in pandas read_csv, to keep importing as much data as possible from a text file in the presence of errors.