通过WCF扩展实现消息压缩

对于需要进行大规模数据传输的WCF应用来说，对于请求消息和回复消息进行传输前的压缩，不但可以降低网络流量，也可以提高网络传输的性能。由于WCF的扩展性，我们可以采用不同的方式实现对消息的压缩，本文提供一种比较简单的实现方式。[源代码从这里下载]

一、三种可行的消息压缩方案
二、DataCompressor——用于数据压缩与解压缩组件
三、MessageCompressor——用于消息压缩与解压的组件
四、CompressionMessageFormatter——用于对请求/回复消息压缩和解压缩的组件
五、CompressionOperationBehaviorAttribute——将CompressionMessageFormatter用于WCF运行时框架的操作行为
六、查看结构压缩后的消息
七、补充说明

一、三种可行的消息压缩方案

消息压缩在WCF中的实现其实很简单，我们只需要在消息（请求消息/回复消息）被序列化之后，发送之前进行压缩；在接收之后，反序列化之前进行解压缩即可。针对压缩/解压缩使用的时机，我们具有三种典型的解决方案。

• 通过自定义MessageEncoder和MessageEncodingBindingElement 来完成。具体的实现，可以参阅张玉彬的文章《WCF进阶:将编码后的字节流压缩传输》和MSDN的文章《Custom Message Encoder: Compression Encoder》。
• 直接创建用于压缩和解压缩的信道，在CodePlex中具有这么一个WCF Extensions；
• 自定义MessageFormatter实现序列化后的压缩和法序列化前的解压缩，这就是我们今天将要介绍的解决方案。

二、DataCompressor——用于数据压缩与解压缩组件

我们支持两种方式的压缩，Dflate和GZip。两种不同的压缩算法通过如下定义的CompressionAlgorithm枚举表示。

public enum CompressionAlgorithm
{
    GZip,
    Deflate
}

而如下定义的DataCompressor负责基于上述两种压缩算法实际上的压缩和解压缩工作。

internal class DataCompressor
{
    public static byte[] Compress(byte[] decompressedData, CompressionAlgorithm algorithm)
    {
        using (MemoryStream stream = new MemoryStream())
        {
            if (algorithm == CompressionAlgorithm.Deflate)
            {
                GZipStream stream2 = new GZipStream(stream, CompressionMode.Compress, true);
                stream2.Write(decompressedData, 0, decompressedData.Length);
                stream2.Close();
            }
            else
            {
                DeflateStream stream3 = new DeflateStream(stream, CompressionMode.Compress, true);
                stream3.Write(decompressedData, 0, decompressedData.Length);
                stream3.Close();
            }
            return stream.ToArray();
        }
    }

    public static byte[] Decompress(byte[] compressedData, CompressionAlgorithm algorithm)
    {
        using (MemoryStream stream = new MemoryStream(compressedData))
        {
            if (algorithm == CompressionAlgorithm.Deflate)
            {
                using (GZipStream stream2 = new GZipStream(stream, CompressionMode.Decompress))
                {
                    return LoadToBuffer(stream2);
                }
            }
            else
            {
                using (DeflateStream stream3 = new DeflateStream(stream, CompressionMode.Decompress))
                {
                    return LoadToBuffer(stream3);
                }
            }
        }
    }

    private static byte[] LoadToBuffer(Stream stream)
    {
        using (MemoryStream stream2 = new MemoryStream())
        {
            int num;
            byte[] buffer = new byte[0x400];
            while ((num = stream.Read(buffer, 0, buffer.Length)) > 0)
            {
                stream2.Write(buffer, 0, num);
            }
            return stream2.ToArray();
        }
    }
}

三、MessageCompressor——用于消息压缩与解压的组件

而针对消息的压缩和解压缩通过如下一个MessageCompressor来完成。具体来说，我们通过上面定义的DataCompressor对消息的主体部分内容进行压缩，并将压缩后的内容存放到一个预定义的XML元素中（名称和命名空间分别为CompressedBody和http://www.artech.com/comporession/），同时添加相应的MessageHeader表示消息经过了压缩，以及采用的压缩算法。对于解压缩，则是通过消息是否具有相应的MessageHeader判断该消息是否经过压缩，如果是则根据相应的算法对其进行解压缩。具体的实现如下：

public class MessageCompressor
 {
     public MessageCompressor(CompressionAlgorithm algorithm)
     {
         this.Algorithm = algorithm;
     }
     public Message CompressMessage(Message sourceMessage)
     {
         byte[] buffer;
         using (XmlDictionaryReader reader1 = sourceMessage.GetReaderAtBodyContents())
         {
             buffer = Encoding.UTF8.GetBytes(reader1.ReadOuterXml());
         }
         if (buffer.Length == 0)
         {
             Message emptyMessage = Message.CreateMessage(sourceMessage.Version, (string)null);
             sourceMessage.Headers.CopyHeadersFrom(sourceMessage);
             sourceMessage.Properties.CopyProperties(sourceMessage.Properties);
             emptyMessage.Close();
             return emptyMessage;
         }
         byte[] compressedData = DataCompressor.Compress(buffer, this.Algorithm);
         string copressedBody = CompressionUtil.CreateCompressedBody(compressedData);
         XmlTextReader reader = new XmlTextReader(new StringReader(copressedBody), new NameTable());
         Message message2 = Message.CreateMessage(sourceMessage.Version, null, (XmlReader)reader);
         message2.Headers.CopyHeadersFrom(sourceMessage);
         message2.Properties.CopyProperties(sourceMessage.Properties);
         message2.AddCompressionHeader(this.Algorithm);
         sourceMessage.Close();
         return message2;
     }

     public Message DecompressMessage(Message sourceMessage)
     {
         if (!sourceMessage.IsCompressed())
         {
             return sourceMessage;
         }
         CompressionAlgorithm algorithm = sourceMessage.GetCompressionAlgorithm();
         sourceMessage.RemoveCompressionHeader();
         byte[] compressedBody = sourceMessage.GetCompressedBody();
         byte[] decompressedBody = DataCompressor.Decompress(compressedBody, algorithm);
         string newMessageXml = Encoding.UTF8.GetString(decompressedBody);
         XmlTextReader reader2 = new XmlTextReader(new StringReader(newMessageXml));
         Message newMessage = Message.CreateMessage(sourceMessage.Version, null, reader2);
         newMessage.Headers.CopyHeadersFrom(sourceMessage);
         newMessage.Properties.CopyProperties(sourceMessage.Properties);
         return newMessage;
     }

     public CompressionAlgorithm Algorithm { get; private set; }
 }

下面是针对Message类型而定义了一些扩展方法和辅助方法。

public static class CompressionUtil
{
    public const string CompressionMessageHeader = "Compression";
    public const string CompressionMessageBody = "CompressedBody";
    public const string Namespace = "http://www.artech.com/compression";

    public static bool IsCompressed(this Message message)
    {
        return message.Headers.FindHeader(CompressionMessageHeader, Namespace) > -1;
    }

    public static void AddCompressionHeader(this Message message, CompressionAlgorithm algorithm)
    {
        message.Headers.Add(MessageHeader.CreateHeader(CompressionMessageHeader, Namespace, string.Format("algorithm = \"{0}\"",algorithm)));
    }

    public static void RemoveCompressionHeader(this Message message)
    {
        message.Headers.RemoveAll(CompressionMessageHeader, Namespace);
    }

    public static CompressionAlgorithm GetCompressionAlgorithm(this Message message)
    {
        if (message.IsCompressed())
        {
            var algorithm = message.Headers.GetHeader<string>(CompressionMessageHeader, Namespace);
            algorithm = algorithm.Replace("algorithm =", string.Empty).Replace("\"", string.Empty).Trim();
            if (algorithm == CompressionAlgorithm.Deflate.ToString())
            {
                return CompressionAlgorithm.Deflate;
            }

            if (algorithm == CompressionAlgorithm.GZip.ToString())
            {
                return CompressionAlgorithm.GZip;
            }
            throw new InvalidOperationException("Invalid compression algrorithm!");
        }
        throw new InvalidOperationException("Message is not compressed!");
    }

    public static byte[] GetCompressedBody(this Message message)
    {
        byte[] buffer;
        using (XmlReader reader1 = message.GetReaderAtBodyContents())
        {
            buffer = Convert.FromBase64String(reader1.ReadElementString(CompressionMessageBody, Namespace));
        }
        return buffer;
    }

    public static string CreateCompressedBody(byte[] content)
    {
        StringWriter output = new StringWriter();
        using (XmlWriter writer2 = XmlWriter.Create(output))
        {
            writer2.WriteStartElement(CompressionMessageBody, Namespace);
            writer2.WriteBase64(content, 0, content.Length);
            writer2.WriteEndElement();
        }
        return output.ToString();
    }
}

四、CompressionMessageFormatter——用于对请求/回复消息压缩和解压缩的组件

消息的序列化和反序列化最终是通过MessageFormatter来完成的。具体来说，客户端通过ClientMessageFormatter实现对请求消息的序列化和对回复消息的序列化，而服务端通过DispatchMessageFormatter实现对请求消息的反序列化和对回复消息的序列化。

在默认的情况下，WCF选用的MessageFormatter为DataContractSerializerOperationFormatter，它采用DataContractSerializer进行实际的序列化和法序列化操作。我们自定义的MessageFormatter实际上是对DataContractSerializerOperationFormatter的封装，我们依然使用它来完成序列化和反序列化工作，额外实现序列化后的压缩和法序列化前的解压缩。

因为DataContractSerializerOperationFormatter是一个internal类型，我们只有通过反射的方式来创建它。如下的代码片断为用于进行消息压缩与解压缩的自定义MessageFormatter，即CompressionMessageFormatter的定义。

public class CompressionMessageFormatter: IDispatchMessageFormatter, IClientMessageFormatter
{
    private const string DataContractSerializerOperationFormatterTypeName = "System.ServiceModel.Dispatcher.DataContractSerializerOperationFormatter, System.ServiceModel, Version=4.0.0.0, Culture=neutral, PublicKeyToken=b77a5c561934e089";

    public IDispatchMessageFormatter InnerDispatchMessageFormatter { get; private set; }
    public IClientMessageFormatter InnerClientMessageFormatter { get; private set; }
    public MessageCompressor MessageCompressor { get; private set; }

    public CompressionMessageFormatter(CompressionAlgorithm algorithm, OperationDescription description, DataContractFormatAttribute dataContractFormatAttribute, DataContractSerializerOperationBehavior serializerFactory)
    {
        this.MessageCompressor = new MessageCompressor(algorithm);
        Type innerFormatterType = Type.GetType(DataContractSerializerOperationFormatterTypeName);
        var innerFormatter = Activator.CreateInstance(innerFormatterType, description, dataContractFormatAttribute, serializerFactory);
        this.InnerClientMessageFormatter = innerFormatter as IClientMessageFormatter;
        this.InnerDispatchMessageFormatter = innerFormatter as IDispatchMessageFormatter;
    }

    public void DeserializeRequest(Message message, object[] parameters)
    {
        message = this.MessageCompressor.DecompressMessage(message);
        this.InnerDispatchMessageFormatter.DeserializeRequest(message, parameters);
    }

    public Message SerializeReply(MessageVersion messageVersion, object[] parameters, object result)
    {
        var message = this.InnerDispatchMessageFormatter.SerializeReply(messageVersion, parameters, result);
        return this.MessageCompressor.CompressMessage(message);
    }

    public object DeserializeReply(Message message, object[] parameters)
    {
        message = this.MessageCompressor.DecompressMessage(message);
        return this.InnerClientMessageFormatter.DeserializeReply(message, parameters);
    }

    public Message SerializeRequest(MessageVersion messageVersion, object[] parameters)
    {
        var message = this.InnerClientMessageFormatter.SerializeRequest(messageVersion, parameters);
        return this.MessageCompressor.CompressMessage(message);
    }
}

五、CompressionOperationBehaviorAttribute——将CompressionMessageFormatter用于WCF运行时框架的操作行为

ClientMessageFormatter和DispatchMessageFormatter实际上属于ClientOperation和DispatchOperation的组件。我们可以通过如下一个自定义的操作行为CompressionOperationBehaviorAttribute将其应用到相应的操作上。

[AttributeUsage( AttributeTargets.Method)]
public class CompressionOperationBehaviorAttribute: Attribute, IOperationBehavior
{
    public CompressionAlgorithm Algorithm { get; set; }

    public void AddBindingParameters(OperationDescription operationDescription, BindingParameterCollection bindingParameters) { }

    public void ApplyClientBehavior(OperationDescription operationDescription, ClientOperation clientOperation)
    {
        clientOperation.SerializeRequest = true;
        clientOperation.DeserializeReply = true;
        var dataContractFormatAttribute = operationDescription.SyncMethod.GetCustomAttributes(typeof(DataContractFormatAttribute), true).FirstOrDefault() as DataContractFormatAttribute;
        if (null == dataContractFormatAttribute)
        {
            dataContractFormatAttribute = new DataContractFormatAttribute();
        }

        var dataContractSerializerOperationBehavior = operationDescription.Behaviors.Find<DataContractSerializerOperationBehavior>();
        clientOperation.Formatter = new CompressionMessageFormatter(this.Algorithm, operationDescription, dataContractFormatAttribute, dataContractSerializerOperationBehavior);
    }

    public void ApplyDispatchBehavior(OperationDescription operationDescription, DispatchOperation dispatchOperation)
    {
        dispatchOperation.SerializeReply        = true;
        dispatchOperation.DeserializeRequest    = true;
        var dataContractFormatAttribute = operationDescription.SyncMethod.GetCustomAttributes(typeof(DataContractFormatAttribute), true).FirstOrDefault() as DataContractFormatAttribute;
        if (null == dataContractFormatAttribute)
        {
            dataContractFormatAttribute = new DataContractFormatAttribute();
        }
        var dataContractSerializerOperationBehavior = operationDescription.Behaviors.Find<DataContractSerializerOperationBehavior>();
        dispatchOperation.Formatter = new CompressionMessageFormatter(this.Algorithm, operationDescription, dataContractFormatAttribute, dataContractSerializerOperationBehavior);
    }

    public void Validate(OperationDescription operationDescription) { }
}

六、查看结构压缩后的消息

为了验证应用了CompressionOperationBehaviorAttribute特性的操作方法对应的消息是否经过了压缩，我们可以通过一个简单的例子来检验。我们采用常用的计算服务的例子，下面是服务契约和服务类型的定义。我们上面定义的CompressionOperationBehaviorAttribute应用到服务契约的Add操作上。

[ServiceContract(Namespace= "http://www.artech.com/")]
public interface ICalculator
{
    [OperationContract]
    [CompressionOperationBehavior]
    double Add(double x, double y);
}
public class CalculatorService : ICalculator
{
    public double Add(double x, double y)
    {
        return x + y;
    }
}

我们采用BasicHttpBinding作为终结点的绑定类型（具体的配置请查看源代码），下面是通过Fiddler获取的消息的内容，它们的主体部分都经过了基于压缩的编码。

<s:Envelope xmlns:s="http://schemas.xmlsoap.org/soap/envelope/">
  <s:Header>
    <Compression xmlns="http://www.artech.com/compression">algorithm = "GZip"</Compression>
  </s:Header>
  <s:Body>
    <CompressedBody xmlns="http://www.artech.com/compression">7L0HYBx ... CQAA//8=</CompressedBody>
  </s:Body>
</s:Envelope>

回复消息

<s:Envelope xmlns:s="http://schemas.xmlsoap.org/soap/envelope/">
  <s:Header>
    <Compression xmlns="http://www.artech.com/compression">algorithm = "GZip"</Compression>
  </s:Header>
  <s:Body>
    <CompressedBody xmlns="http://www.artech.com/compression">7L0H...PAAAA//8=</CompressedBody>
  </s:Body>
</s:Envelope>

七、补充说明

由于CompressionMessageFormatter使用基于DataContractSerializer序列化器的DataContractSerializerOperationFormatter进行消息的序列化和发序列化工作。而DataContractSerializer仅仅是WCF用于序列化的一种默认的选择（WCF还可以采用传统的XmlSeriaizer）。为了让CompressionMessageFormatter能够使用其他序列化器，你可以对于进行相应的修正。