ZStd for .NET

ZStd is a multi-platform .NET binding of Facebook's Zstandard library.

Supported platforms:

Windows
- x64
- x86
Linux
- GNU flavors (Debian, Ubuntu, ...)
- Musl flavors (Alpine, Void, ...)
- Intel x86
- AMD64 / Intel x86-64
- ARMv8-64 / AArch64
- ARMv7-HF / AArch32
Apple OSX
- 64-bit only

Note: ARM support is currently hard-float only, no soft-float platform support unless requested.

Note: In the not too distant future support for 32-bit architectures will likely be dropped.

Features:

Compression and decompression
- Dictionary support
- Managed streams into streams of frames
- Byte arrays and spans into frames
- Byte arrays and spans into blocks (without frames)
Generation of dictionaries from a collection of samples
Loading and saving of dictionaries to and from streams, arrays and spans
Helpers for collection of dictionary samples and training
Training will automatically tune unspecified dictionary parameters

Take a look at the unit tests to explore its behavior in different situations.

This is in the later stages of development, and features are less likely to be added over time.

Known Issues:

Currently the native dependencies are shipped with this NuGet package for all platforms. Separate NuGet runtime packages should be created to provide each specific platform dependency.
Coverage currently stands at 50%. Critical path coverage is high. Coverage of non-critical path operations is low.
The GC is not yet automatically made aware of unmanaged memory usage.

Acknowlegedments

Official documentation

This project is heavily inspired by ZstdNet, but shares very little in the way of implementation.

Reference

Basic usage examples

Dictionaries

// pick a compression level or use 0, considered 'auto' in most cases
var compressionLevel = 3;
//var compressionLevel = ZStdCompressor.MinimumCompressionLevel; // probably 1
// Note: negative levels do exist as per zstd documentation, but this just calls ZSTD_minCLevel
//var compressionLevel = ZStdCompressor.MaximumCompressionLevel; // probably 22

// allocate a 32kB buffer to build a new dictionary
var dict = new ZStdDictionaryBuilder(32 * 1024);
// or load from some external data
//var dict = new ZStdDictionaryBuilder(someBytes);
//var dict = new ZStdDictionaryBuilder(someBytes, someSize);


// train the dictionary, and retrieve tuned training parameters for future training
var trainedParams = dict.Train(async () => {
  // sample your data by returning ArraySegment<byte>s of regions you expect to occur often
  // refer to Facebook's Zstd documentation for details on what should be sampled
  yield return ArraySegment<byte>(trainingData, 7, 42);
}, compressionLevel);

// optionally save the dictionary to a file somehow
using (var fs = File.Create(path))
  dict.WriteTo(fs); // preferred
//  fs.Write(dict); // implicitly casts to ReadOnlySpan<byte>
//File.WriteAllBytes("saved_dictionary", ((ArraySegment<byte>)dict).ToArray());

// create an unmanaged reference for use with compression
using var cDict = dict.CreateCompressorDictionary(compressionLevel);

// create an unmanaged reference for use with decompression
using var dDict = dict.CreateDecompressorDictionary();

// be sure to dispose of your compressors/decompressors before your dictionary references

Compression

// create a context
using var cCtx = new ZStdCompressor();

// specify the dictionary you want to use, or don't
cCtx.UseDictionary(cDict);

// figure out about how big of a buffer you need
var compressBufferSize = CCtx.GetUpperBound((UIntPtr) inputData.Length);

var compressBuffer = new byte[compressBufferSize];

// set the compressor to the compression level (it may inherit it from the dictionary)
cCtx.Set(CompressionParameter.CompressionLevel, compressionLevel);

// actually perform the compression operation
var compressedSize = cCtx.Compress(compressBuffer, inputData);

// retrieve your compressed frame
var compressedFrame = new ArraySegment<byte>(compressBuffer, 0, (int) compressedSize);

Decompression

// create a context
var dCtx = new ZStdDecompressor();

// specify the dictionary you want to use, or don't
dCtx.UseDictionary(dDict);

// figure out about how big of a buffer you need
var decompressBufferSize = DCtx.GetUpperBound(compressedFrame);

var decompressBuffer = new byte[decompressBufferSize];

// actually perform the decompression operation
var decompressedSize = dCtx.Decompress(decompressBuffer, compressedFrame);

// retrieve your decompressed frame
var decompressedFrame = new ArraySegment<byte>(decompressBuffer, 0, (int) decompressedSize);

Stream Compression

// have somewhere to put the compressed output, any stream will do
using var compressed = new MemoryStream();

// create a compression context over the output stream
using var compressStream = new ZStdCompressStream(compressed);

// ZStdCompressStream.Compressor is just a ZStdCompressor, so you can set parameters and use dictionaries
compressStream.Compressor.Set(CompressionParameter.CompressionLevel, compressionLevel);

// write some data into it
compressStream.Write(someData);

// if you're done writing data and you're not at the end of a frame, it'd be wise to flush
compressStream.Flush();

Stream Decompression

// consume the compressed input stream with the stream decompressor
using var decompressStream = new ZStdDecompressStream(compressedInput);

// you can access ZStdCompressStream.Decompressor, it's a ZStdDecmpressor, to specify a dictionary and such

// decompress some data..
var decompressed = new byte[someDataSize];

decompressStream.Read(decompressed, 0, decompressed.Length);

ImpromptuNinjas/ZStd