ZeroBit - A high-performance file compressor written in modern C++, leveraging a multi-stage pipeline-including Burrows-Wheeler Transform (BWT), Move-To-Front (MTF), Zero Run-Length Encoding (RLE), and an adaptive context-mixing range coder-to achieve efficient, lossless compression. Specifically created for small files.
- Bidirectional: Supports both compression and decompression.
- Adaptive Modeling: Combines multiple context models (byte, bit, match, LZP) with online mixing.
- Single-header Implementation: Minimal dependencies; requires only C++17 and Qt 6.9.0.
- Portable: Uses
std::filesystemfor cross-platform file handling.
- Best for: Text files with a lot of numerical values.
- Bad for: Text files with a lot of non-structured text.
Each compressed file's header is structured as follows:
- Original file size (uint64_t)
- Block length (uint32_t)
- BWT primary index (uint32_t)
- RLE symbol count (uint32_t)
- Compressed data size (uint32_t)
- Range-coded payload (bytes)
This design enables streaming decompression without loading the entire file into memory.
| Stage | Technique | Purpose |
|---|---|---|
| 1 | Burrows–Wheeler Transform (BWT) | Increases symbol locality |
| 2 | Move-To-Front (MTF) | Exposes runs of low symbols |
| 3 | Zero Run-Length Encoding (RLE) | Efficiently encodes zero runs |
| 4 | Adaptive Context Models + Mixer | Learns bitwise patterns dynamically |
| 5 | Range Coding | Optimal entropy encoding |
Developed by Stefan Rusev (stefanrusev02@gmail.com)