XmlDictionaryReader 读取固定大小的归零流

有人能给我很好的解释为什么这会失败吗？

const int bufferSize = 2 * 1024, testValue = 123456;
var buffer = new byte[bufferSize];

var serializer = new DataContractSerializer(typeof(int));

//Serialize value
using (var memStream = new MemoryStream(buffer))
using (XmlDictionaryWriter writer = XmlDictionaryWriter.CreateBinaryWriter(memStream))
    serializer.WriteObject(writer, testValue);

//Deserialize value
using (var memStream = new MemoryStream(buffer))
using (XmlDictionaryReader reader = XmlDictionaryReader.CreateBinaryReader(memStream, XmlDictionaryReaderQuotas.Max))
{
    object deserializedValue = serializer.ReadObject(reader); // <--- nope, this throws System.Runtime.Serialization.SerializationException: The input source is not correctly formatted.
    Console.WriteLine(deserializedValue);
}

我正在使用共享内存 IPC，其中一部分意味着您必须处理固定大小的进程间缓冲区。我正在将对象序列化到缓冲区中，并且我想变得很酷，所以我尝试使用 DataContractSerializer + 二进制 XmlDictionaryWriter 组合，这是我所知道的最快的非自定义序列化技术之一。 问题是，反序列化时，XmlDictionaryReader 似乎试图将整个内存流视为一个大的 xml 文档，并读取它自己的流结束/块标记，遇到大堆零，然后简单地自行崩溃。 BinaryFormatter 不存在这个问题，因为它逐块读取流。 我不得不想出一个相当蹩脚的解决方案来实现自定义流，该解决方案在到达第一个 0 后“伪造”流的结尾（假设是 XmlDictionaryWriter 的 eof 标记）。

完整演示：

using System;
using System.IO;
using System.Runtime.Serialization;
using System.Xml;

namespace SerializationTest
{
    public static class Program
    {
        public static void Main()
        {
            const int bufferSize = 2 * 1024, testValue = 123456;
            var buffer = new byte[bufferSize];

            var serializer = new DataContractSerializer(typeof(int));

            //Serialize value
            using (var memStream = new MemoryStream(buffer))
            using (XmlDictionaryWriter writer = XmlDictionaryWriter.CreateBinaryWriter(memStream))
                serializer.WriteObject(writer, testValue);

            //Deserialize value
            using (var memStream = new MemoryStream(buffer))
            using (XmlDictionaryReader reader = XmlDictionaryReader.CreateBinaryReader(memStream, XmlDictionaryReaderQuotas.Max))
            {
                object deserializedValue = serializer.ReadObject(reader); // <--- nope, this throws System.Runtime.Serialization.SerializationException: The input source is not correctly formatted.
                Console.WriteLine(deserializedValue);
            }

            //Deserialize value via FakeEndStream
            using (var memStream = new FakeEndStream(new MemoryStream(buffer)))
            using (XmlDictionaryReader reader = XmlDictionaryReader.CreateBinaryReader(memStream, XmlDictionaryReaderQuotas.Max))
            {
                object deserializedValue = serializer.ReadObject(reader);
                Console.WriteLine(deserializedValue);
            }
        }

        private sealed class FakeEndStream : Stream
        {
            private readonly Stream _source;
            private bool _endOfStream;

            public FakeEndStream(Stream source)
            {
                _source = source;
            }

            #region The workaround

            public override int Read(byte[] buffer, int offset, int count)
            {
                int i = 0;

                for (int position = offset; i < count; i++, position++)
                {
                    int value = ReadByte();
                    if (value < 0)
                        return i;

                    buffer[position] = (byte)value;
                }

                return i;
            }

            public override int ReadByte()
            {
                if (_endOfStream)
                    return -1;

                int value = _source.ReadByte();

                if (value <= 0)
                    _endOfStream = true;

                return value;
            }

            #endregion

            #region Boilerplate overrides of Stream

            protected override void Dispose(bool disposing)
            {
                _source.Dispose();
            }

            public override void Flush()
            {
                _source.Flush();
            }

            public override long Seek(long offset, SeekOrigin origin)
            {
                return _source.Seek(offset, origin);
            }

            public override void SetLength(long value)
            {
                _source.SetLength(value);
            }

            public override void Write(byte[] buffer, int offset, int count)
            {
                _source.Write(buffer, offset, count);
            }

            public override bool CanRead
            {
                get { return _source.CanRead; }
            }

            public override bool CanSeek
            {
                get { return _source.CanSeek; }
            }

            public override bool CanWrite
            {
                get { return _source.CanWrite; }
            }

            public override long Length
            {
                get { return _source.Length; }
            }

            public override long Position
            {
                get { return _source.Position; }
                set { _source.Position = value; }
            }

            #endregion
        }
    }
}

当然，我会帮助你解决这个问题（因为没有人愿意去确认/否认这些症状）。

我发现问题确实是读者读取了无效数据。但是，您可以通过在流末尾写入空格来帮助它，如下所示：

//Serialize value
using (var memStream = new MemoryStream(buffer))
using (XmlDictionaryWriter writer = XmlDictionaryWriter.CreateBinaryWriter(memStream))
{
    serializer.WriteObject(writer, testValue);
    writer.WriteWhitespace(" ");
}

//Deserialize value
using (var memStream = new MemoryStream(buffer))
using (XmlDictionaryReader reader = XmlDictionaryReader.CreateBinaryReader(memStream, XmlDictionaryReaderQuotas.Max))
{
    object deserializedValue = serializer.ReadObject(reader); // \o/
    Console.WriteLine(deserializedValue);
}

不敢相信没有人喜欢这个问题，因为它可能会导致运行时难以调试异常，从而使使用 XmlDictionaryWriter 序列化对象变得有点不可靠。

连接报告：https://connect.microsoft.com/VisualStudio/feedback/details/811170/xmlbinaryreader-not-able-to-read-from-fixed-size-buffer https://connect.microsoft.com/VisualStudio/feedback/details/811170/xmlbinaryreader-not-able-to-read-from-fixed-size-buffer