packrat

A compression tool that works well on text and source code.

Uses the BWT-MTF-RLE-Huffman pipeline (same approach as bzip2). Achieves around 10-20% ratios on typical source code.

Build

make

Usage

Compress a file:

./packrat -c myfile.txt myfile.prt
./packrat -c myfile.txt              # auto-names to myfile.txt.prt

Decompress:

./packrat -d myfile.prt output.txt
./packrat -d myfile.txt.prt          # auto-names to myfile.txt

Create an archive:

./packrat -a archive.prt src/

Create a solid archive (better compression, files grouped by type):

./packrat -a --solid archive.prt src/

Extract:

./packrat -x archive.prt dest/

List contents:

./packrat -l archive.prt

Show file info:

./packrat -i compressed.prt

How it works

1. BWT - Burrows-Wheeler Transform groups similar characters together
2. MTF - Move-to-Front converts those clusters into runs of small numbers
3. RLE - Run-length encoding compresses the zero runs
4. Huffman - Entropy coding assigns short codes to common symbols

Solid archives concatenate files by type before compressing. All your .c files get compressed together, all your .py files together, etc. This exploits cross-file redundancy.

File format

• .prt extension
• Magic bytes: PRT\x02 (single file) or PRT\x03 (solid archive)
• Stores original filename for auto-naming on decompress

Tests

./tests/test_packrat.sh ./packrat

License

MIT