1<p align="center"><img src="https://raw.githubusercontent.com/facebook/zstd/dev/doc/images/zstd_logo86.png" alt="Zstandard"></p> 2 3__Zstandard__, or `zstd` as short version, is a fast lossless compression algorithm, 4targeting real-time compression scenarios at zlib-level and better compression ratios. 5It's backed by a very fast entropy stage, provided by [Huff0 and FSE library](https://github.com/Cyan4973/FiniteStateEntropy). 6 7The project is provided as an open-source dual [BSD](LICENSE) and [GPLv2](COPYING) licensed **C** library, 8and a command line utility producing and decoding `.zst`, `.gz`, `.xz` and `.lz4` files. 9Should your project require another programming language, 10a list of known ports and bindings is provided on [Zstandard homepage](http://www.zstd.net/#other-languages). 11 12**Development branch status:** 13 14[![Build Status][travisDevBadge]][travisLink] 15[![Build status][AppveyorDevBadge]][AppveyorLink] 16[![Build status][CircleDevBadge]][CircleLink] 17[![Build status][CirrusDevBadge]][CirrusLink] 18[![Fuzzing Status][OSSFuzzBadge]][OSSFuzzLink] 19 20[travisDevBadge]: https://travis-ci.org/facebook/zstd.svg?branch=dev "Continuous Integration test suite" 21[travisLink]: https://travis-ci.org/facebook/zstd 22[AppveyorDevBadge]: https://ci.appveyor.com/api/projects/status/xt38wbdxjk5mrbem/branch/dev?svg=true "Windows test suite" 23[AppveyorLink]: https://ci.appveyor.com/project/YannCollet/zstd-p0yf0 24[CircleDevBadge]: https://circleci.com/gh/facebook/zstd/tree/dev.svg?style=shield "Short test suite" 25[CircleLink]: https://circleci.com/gh/facebook/zstd 26[CirrusDevBadge]: https://api.cirrus-ci.com/github/facebook/zstd.svg?branch=dev 27[CirrusLink]: https://cirrus-ci.com/github/facebook/zstd 28[OSSFuzzBadge]: https://oss-fuzz-build-logs.storage.googleapis.com/badges/zstd.svg 29[OSSFuzzLink]: https://bugs.chromium.org/p/oss-fuzz/issues/list?sort=-opened&can=1&q=proj:zstd 30 31## Benchmarks 32 33For reference, several fast compression algorithms were tested and compared 34on a server running Arch Linux (`Linux version 5.0.5-arch1-1`), 35with a Core i9-9900K CPU @ 5.0GHz, 36using [lzbench], an open-source in-memory benchmark by @inikep 37compiled with [gcc] 8.2.1, 38on the [Silesia compression corpus]. 39 40[lzbench]: https://github.com/inikep/lzbench 41[Silesia compression corpus]: http://sun.aei.polsl.pl/~sdeor/index.php?page=silesia 42[gcc]: https://gcc.gnu.org/ 43 44| Compressor name | Ratio | Compression| Decompress.| 45| --------------- | ------| -----------| ---------- | 46| **zstd 1.4.0 -1** | 2.884 | 530 MB/s | 1360 MB/s | 47| zlib 1.2.11 -1 | 2.743 | 110 MB/s | 440 MB/s | 48| brotli 1.0.7 -0 | 2.701 | 430 MB/s | 470 MB/s | 49| quicklz 1.5.0 -1 | 2.238 | 600 MB/s | 800 MB/s | 50| lzo1x 2.09 -1 | 2.106 | 680 MB/s | 950 MB/s | 51| lz4 1.8.3 | 2.101 | 800 MB/s | 4220 MB/s | 52| snappy 1.1.4 | 2.073 | 580 MB/s | 2020 MB/s | 53| lzf 3.6 -1 | 2.077 | 440 MB/s | 930 MB/s | 54 55[zlib]: http://www.zlib.net/ 56[LZ4]: http://www.lz4.org/ 57 58Zstd can also offer stronger compression ratios at the cost of compression speed. 59Speed vs Compression trade-off is configurable by small increments. 60Decompression speed is preserved and remains roughly the same at all settings, 61a property shared by most LZ compression algorithms, such as [zlib] or lzma. 62 63The following tests were run 64on a server running Linux Debian (`Linux version 4.14.0-3-amd64`) 65with a Core i7-6700K CPU @ 4.0GHz, 66using [lzbench], an open-source in-memory benchmark by @inikep 67compiled with [gcc] 7.3.0, 68on the [Silesia compression corpus]. 69 70Compression Speed vs Ratio | Decompression Speed 71---------------------------|-------------------- 72 |  73 74A few other algorithms can produce higher compression ratios at slower speeds, falling outside of the graph. 75For a larger picture including slow modes, [click on this link](doc/images/DCspeed5.png). 76 77 78## The case for Small Data compression 79 80Previous charts provide results applicable to typical file and stream scenarios (several MB). Small data comes with different perspectives. 81 82The smaller the amount of data to compress, the more difficult it is to compress. This problem is common to all compression algorithms, and reason is, compression algorithms learn from past data how to compress future data. But at the beginning of a new data set, there is no "past" to build upon. 83 84To solve this situation, Zstd offers a __training mode__, which can be used to tune the algorithm for a selected type of data. 85Training Zstandard is achieved by providing it with a few samples (one file per sample). The result of this training is stored in a file called "dictionary", which must be loaded before compression and decompression. 86Using this dictionary, the compression ratio achievable on small data improves dramatically. 87 88The following example uses the `github-users` [sample set](https://github.com/facebook/zstd/releases/tag/v1.1.3), created from [github public API](https://developer.github.com/v3/users/#get-all-users). 89It consists of roughly 10K records weighing about 1KB each. 90 91Compression Ratio | Compression Speed | Decompression Speed 92------------------|-------------------|-------------------- 93 |  |  94 95 96These compression gains are achieved while simultaneously providing _faster_ compression and decompression speeds. 97 98Training works if there is some correlation in a family of small data samples. The more data-specific a dictionary is, the more efficient it is (there is no _universal dictionary_). 99Hence, deploying one dictionary per type of data will provide the greatest benefits. 100Dictionary gains are mostly effective in the first few KB. Then, the compression algorithm will gradually use previously decoded content to better compress the rest of the file. 101 102### Dictionary compression How To: 103 1041. Create the dictionary 105 106 `zstd --train FullPathToTrainingSet/* -o dictionaryName` 107 1082. Compress with dictionary 109 110 `zstd -D dictionaryName FILE` 111 1123. Decompress with dictionary 113 114 `zstd -D dictionaryName --decompress FILE.zst` 115 116 117## Build instructions 118 119### Makefile 120 121If your system is compatible with standard `make` (or `gmake`), 122invoking `make` in root directory will generate `zstd` cli in root directory. 123 124Other available options include: 125- `make install` : create and install zstd cli, library and man pages 126- `make check` : create and run `zstd`, tests its behavior on local platform 127 128### cmake 129 130A `cmake` project generator is provided within `build/cmake`. 131It can generate Makefiles or other build scripts 132to create `zstd` binary, and `libzstd` dynamic and static libraries. 133 134By default, `CMAKE_BUILD_TYPE` is set to `Release`. 135 136### Meson 137 138A Meson project is provided within [`build/meson`](build/meson). Follow 139build instructions in that directory. 140 141You can also take a look at [`.travis.yml`](.travis.yml) file for an 142example about how Meson is used to build this project. 143 144Note that default build type is **release**. 145 146### Visual Studio (Windows) 147 148Going into `build` directory, you will find additional possibilities: 149- Projects for Visual Studio 2005, 2008 and 2010. 150 + VS2010 project is compatible with VS2012, VS2013, VS2015 and VS2017. 151- Automated build scripts for Visual compiler by [@KrzysFR](https://github.com/KrzysFR), in `build/VS_scripts`, 152 which will build `zstd` cli and `libzstd` library without any need to open Visual Studio solution. 153 154### Buck 155 156You can build the zstd binary via buck by executing: `buck build programs:zstd` from the root of the repo. 157The output binary will be in `buck-out/gen/programs/`. 158 159## Status 160 161Zstandard is currently deployed within Facebook. It is used continuously to compress large amounts of data in multiple formats and use cases. 162Zstandard is considered safe for production environments. 163 164## License 165 166Zstandard is dual-licensed under [BSD](LICENSE) and [GPLv2](COPYING). 167 168## Contributing 169 170The "dev" branch is the one where all contributions are merged before reaching "master". 171If you plan to propose a patch, please commit into the "dev" branch, or its own feature branch. 172Direct commit to "master" are not permitted. 173For more information, please read [CONTRIBUTING](CONTRIBUTING.md). 174
