mongoimport currently reads json, csv, and tsv files line-by-line. This leads to many small reads and thus poor performance when importing large amounts of data. The tool should be refactored to read large chunks of input and then process these large chunks line-by-line.