Compression results
bbz is a block-sorting compressor based on the Burrows–Wheeler transformation, developed as
part of the infinity package. The tables show bbz -9 (the strongest level, version 2.2) on
standard corpora, compared with gzip -9.
| Corpus | Files | Size | bbz -9 | bits/byte | gzip -9 | bits/byte |
|---|---|---|---|---|---|---|
| Calgary corpus | 14 | 3,141,622 | 744,813 | 2.177 (1.897) | 1,017,624 | 2.695 (2.591) |
| Canterbury corpus | 14 | 13,970,266 | 2,601,234 | 1.868 (1.490) | 3,927,856 | 2.483 (2.249) |
| Protein corpus | 4 | 7,154,401 | 1,988,543 | 1.888 (2.224) | 4,134,309 | 4.626 (4.623) |
Bits per byte: average of the per-file rates, in parentheses the rate over the whole corpus. Each file is compressed as a single block; all results were verified by decompression.