Skip to content

Latest commit

 

History

History
115 lines (84 loc) · 4.49 KB

README.md

File metadata and controls

115 lines (84 loc) · 4.49 KB

Snappier

Introduction

Snappier is a pure C# port of Google's Snappy compression algorithm. It is designed with speed as the primary goal, rather than compression ratio, and is ideal for compressing network traffic. Please see the Snappy README file for more details on Snappy.

Complete documentation is available at https://brantburnett.github.io/Snappier/.

Project Goals

The Snappier project aims to meet the following needs of the .NET community.

  • Cross-platform C# implementation for Linux and Windows, without P/Invoke or special OS installation requirements
  • Compatible with .NET 4.6.1 and later and .NET 6 and later
  • Use .NET paradigms, including asynchronous stream support
  • Full compatibility with both block and stream formats
  • Near C++ level performance
  • Keep allocations and garbage collection to a minimum using buffer pools

Installing

Simply add a NuGet package reference to the latest version of Snappier.

<PackageReference Include="Snappier" Version="1.1.6" />

or

dotnet add package Snappier

Block compression/decompression using a memory pool buffer

using Snappier;

public class Program
{
    private static byte[] Data = {0, 1, 2}; // Wherever you get the data from

    public static void Main()
    {
        // This option uses `MemoryPool<byte>.Shared`. However, if you fail to
        // dispose of the returned buffers correctly it can result in inefficient garbage collection.
        // It is important to either call .Dispose() or use a using statement.

        // Compression
        using (IMemoryOwner<byte> compressed = Snappy.CompressToMemory(Data))
        {
            // Decompression
            using (IMemoryOwner<byte> decompressed = Snappy.DecompressToMemory(compressed.Memory.Span))
            {
                // Do something with the data
            }
        }
    }
}

Stream compression/decompression

Compressing or decompressing a stream follows the same paradigm as other compression streams in .NET. SnappyStream wraps an inner stream. If decompressing you read from the SnappyStream, if compressing you write to the SnappyStream

This approach reads or writes the Snappy framing format designed for streaming. The input/output is not the same as the block method above. It includes additional headers and CRC32C checks.

using System.IO;
using System.IO.Compression;
using Snappier;

public class Program
{
    public static async Task Main()
    {
        using var fileStream = File.OpenRead("somefile.txt");

        // First, compression
        using var compressed = new MemoryStream();

        using (var compressor = new SnappyStream(compressed, CompressionMode.Compress, leaveOpen: true))
        {
            await fileStream.CopyToAsync(compressor);

            // Disposing the compressor also flushes the buffers to the inner stream
            // We pass true to the constructor above so that it doesn't close/dispose the inner stream
            // Alternatively, we could call compressor.Flush()
        }

        // Then, decompression

        compressed.Position = 0; // Reset to beginning of the stream so we can read
        using var decompressor = new SnappyStream(compressed, CompressionMode.Decompress);

        var buffer = new byte[65536];
        var bytesRead = decompressor.Read(buffer, 0, buffer.Length);
        while (bytesRead > 0)
        {
            // Do something with the data

            bytesRead = decompressor.Read(buffer, 0, buffer.Length)
        }
    }
}

Other Projects

There are other projects available for C#/.NET which implement Snappy compression.

  • Snappy.NET - Uses P/Invoke to C++ for great performance. However, it only works on Windows, is a bit heap allocation heavy in some cases, and is a deprecated project. This project may still be the best choice if your project is on the legacy .NET Framework on Windows, where Snappier is much less performant.
  • IronSnappy - Another pure C# port, based on the Golang implementation instead of the C++ implementation.