bzip2

bzip2 is a data compression program and format that uses a block-sorting algorithm to compress data. It was created by Julian Seward and first released in 1996 as part of the bzip family. The compressor works on individual blocks of input data, with block sizes configurable from 100 kilobytes to 900 kilobytes (the default is 900 kilobytes). A .bz2 file consists of a sequence of compressed blocks, each with its own metadata, assembled into a single stream.

On each block, bzip2 applies the Burrows-Wheeler transform to reorder the bytes into a form that is more amenable to compression. This is followed by a move-to-front transform, a run-length encoding step, and finally Huffman coding to produce the bitstream. The blockwise design allows relatively good compression while enabling streaming and random access in some cases, but it also means memory usage and compression time grow with block size.

libbzip2 provides the decompression and compression functionality as a library, and many tools use it to offer .bz2 support. The format is often used to compress tar archives, yielding file names with a .tar.bz2 extension.

In practice, bzip2 generally offers higher compression ratios than gzip for many data types, at the cost of slower compression. Decompression is typically fast, and memory usage scales with the chosen block size. bzip2 has become a standard tool in Unix-like environments and is widely supported across operating systems.