Which sounds ridiculous, but, works around limitations with reading super large CSV files; especially when the map/reduce script is yielded.