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 12Development branch status : [![Build Status][travisDevBadge]][travisLink] [![Build status][AppveyorDevBadge]][AppveyorLink] [![Build status][CircleDevBadge]][CircleLink] 13 14[travisDevBadge]: https://travis-ci.org/facebook/zstd.svg?branch=dev "Continuous Integration test suite" 15[travisLink]: https://travis-ci.org/facebook/zstd 16[AppveyorDevBadge]: https://ci.appveyor.com/api/projects/status/xt38wbdxjk5mrbem/branch/dev?svg=true "Windows test suite" 17[AppveyorLink]: https://ci.appveyor.com/project/YannCollet/zstd-p0yf0 18[CircleDevBadge]: https://circleci.com/gh/facebook/zstd/tree/dev.svg?style=shield "Short test suite" 19[CircleLink]: https://circleci.com/gh/facebook/zstd 20 21### Benchmarks 22 23For reference, several fast compression algorithms were tested and compared 24on a server running Linux Debian (`Linux version 4.14.0-3-amd64`), 25with a Core i7-6700K CPU @ 4.0GHz, 26using [lzbench], an open-source in-memory benchmark by @inikep 27compiled with [gcc] 7.3.0, 28on the [Silesia compression corpus]. 29 30[lzbench]: https://github.com/inikep/lzbench 31[Silesia compression corpus]: http://sun.aei.polsl.pl/~sdeor/index.php?page=silesia 32[gcc]: https://gcc.gnu.org/ 33 34| Compressor name | Ratio | Compression| Decompress.| 35| --------------- | ------| -----------| ---------- | 36| **zstd 1.3.4 -1** | 2.877 | 470 MB/s | 1380 MB/s | 37| zlib 1.2.11 -1 | 2.743 | 110 MB/s | 400 MB/s | 38| brotli 1.0.2 -0 | 2.701 | 410 MB/s | 430 MB/s | 39| quicklz 1.5.0 -1 | 2.238 | 550 MB/s | 710 MB/s | 40| lzo1x 2.09 -1 | 2.108 | 650 MB/s | 830 MB/s | 41| lz4 1.8.1 | 2.101 | 750 MB/s | 3700 MB/s | 42| snappy 1.1.4 | 2.091 | 530 MB/s | 1800 MB/s | 43| lzf 3.6 -1 | 2.077 | 400 MB/s | 860 MB/s | 44 45[zlib]:http://www.zlib.net/ 46[LZ4]: http://www.lz4.org/ 47 48Zstd can also offer stronger compression ratios at the cost of compression speed. 49Speed vs Compression trade-off is configurable by small increments. 50Decompression speed is preserved and remains roughly the same at all settings, 51a property shared by most LZ compression algorithms, such as [zlib] or lzma. 52 53The following tests were run 54on a server running Linux Debian (`Linux version 4.14.0-3-amd64`) 55with a Core i7-6700K CPU @ 4.0GHz, 56using [lzbench], an open-source in-memory benchmark by @inikep 57compiled with [gcc] 7.3.0, 58on the [Silesia compression corpus]. 59 60Compression Speed vs Ratio | Decompression Speed 61---------------------------|-------------------- 62 |  63 64A few other algorithms can produce higher compression ratios at slower speeds, falling outside of the graph. 65For a larger picture including slow modes, [click on this link](doc/images/DCspeed5.png). 66 67 68### The case for Small Data compression 69 70Previous charts provide results applicable to typical file and stream scenarios (several MB). Small data comes with different perspectives. 71 72The 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. 73 74To solve this situation, Zstd offers a __training mode__, which can be used to tune the algorithm for a selected type of data. 75Training 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. 76Using this dictionary, the compression ratio achievable on small data improves dramatically. 77 78The 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). 79It consists of roughly 10K records weighing about 1KB each. 80 81Compression Ratio | Compression Speed | Decompression Speed 82------------------|-------------------|-------------------- 83 |  |  84 85 86These compression gains are achieved while simultaneously providing _faster_ compression and decompression speeds. 87 88Training 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_). 89Hence, deploying one dictionary per type of data will provide the greatest benefits. 90Dictionary 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. 91 92#### Dictionary compression How To: 93 941) Create the dictionary 95 96`zstd --train FullPathToTrainingSet/* -o dictionaryName` 97 982) Compress with dictionary 99 100`zstd -D dictionaryName FILE` 101 1023) Decompress with dictionary 103 104`zstd -D dictionaryName --decompress FILE.zst` 105 106 107### Build instructions 108 109#### Makefile 110 111If your system is compatible with standard `make` (or `gmake`), 112invoking `make` in root directory will generate `zstd` cli in root directory. 113 114Other available options include: 115- `make install` : create and install zstd cli, library and man pages 116- `make check` : create and run `zstd`, tests its behavior on local platform 117 118#### cmake 119 120A `cmake` project generator is provided within `build/cmake`. 121It can generate Makefiles or other build scripts 122to create `zstd` binary, and `libzstd` dynamic and static libraries. 123 124By default, `CMAKE_BUILD_TYPE` is set to `Release`. 125 126#### Meson 127 128A Meson project is provided within `contrib/meson`. 129 130#### Visual Studio (Windows) 131 132Going into `build` directory, you will find additional possibilities: 133- Projects for Visual Studio 2005, 2008 and 2010. 134 + VS2010 project is compatible with VS2012, VS2013, VS2015 and VS2017. 135- Automated build scripts for Visual compiler by [@KrzysFR](https://github.com/KrzysFR), in `build/VS_scripts`, 136 which will build `zstd` cli and `libzstd` library without any need to open Visual Studio solution. 137 138 139### Status 140 141Zstandard is currently deployed within Facebook. It is used continuously to compress large amounts of data in multiple formats and use cases. 142Zstandard is considered safe for production environments. 143 144### License 145 146Zstandard is dual-licensed under [BSD](LICENSE) and [GPLv2](COPYING). 147 148### Contributing 149 150The "dev" branch is the one where all contributions are merged before reaching "master". 151If you plan to propose a patch, please commit into the "dev" branch, or its own feature branch. 152Direct commit to "master" are not permitted. 153For more information, please read [CONTRIBUTING](CONTRIBUTING.md). 154
