Try from_reccords, should be theoretically faster - needs to be tested https://github.com/Haigutus/triplets/blob/main/triplets%2Frdf_parser.py#L438