c# - 缓冲来自套接字的字符串流

Question

我正在开发一个在网络实例之间传递字符串的 C# 应用。

代码目前正在使用异步套接字，到目前为止一切正常(本地主机)。

但是，当我通过真实接口(interface)时，数据包将被拆分和合并时，我预计会出现缓冲问题。

客户:

Socket sock;
// Snip init and connect

string msg1 = "Hello\nWorld";
byte[] data1 = System.Text.Encoding.UTF8.GetBytes(msg1);
sock.Send(data1);

string msg2 = "Foo\nBar\nBaz";
byte[] data2 = System.Text.Encoding.UTF8.GetBytes(msg2);
sock.Send(data2);

我会选择这样的东西，但我想不出一个优雅的解决方案来解决缺失的问题:

服务器:

Socket sock;
MemoryStream ms = new MemoryStream();
Queue<string> strings = new Queue<string>();
// Snip init and receive connection
sock.BeginReceive(buffer, 0, MaxSize, SocketFlags.None, new AsyncCallback(OnReceived), null);

void OnReceived(IAsyncResult result) {
  // Snip sanity stuff
  int bytesReceived = sock.EndReceive(result);

  // Here is where I'd need some help...
  ms.Write(buffer, 0, bytesReceived);
  ms.Flush();
  for (;;) {
    StreamReader sr = new StreamReader(ms);
    if (sr.HasAStringTerminationCharacter) { // <-- How?
      string currentString = sr.ReadUntilTermination(); // <-- How?
      strings.Enqueue(currentString);
    }
    else
      break;
  }
  sock.BeginReceive(buffer, 0, MaxSize, SocketFlags.None, new AsyncCallback(OnReceived), null);
}

Answer 1

由于 Encoding.GetBytes() 不会预先添加或附加字符串终止数据，因此我改用 BinaryReader/BinaryWriter，它在字符串内容之前对字符串长度进行编码。

这是修改后的客户端:

Socket sock;
// Snip init and connect
sock.Send(ToBinary("Hello\nWorld"));
sock.Send(ToBinary("Foo\nBar\nBaz"));

byte[] ToBinary(string s) {
  var ms = new MemoryStream();
  var bw = new BinaryWriter(ms);
  bw.Write(s);
  bw.Flush();
  ms.Flush();
  return ms.ToArray();
}

服务器:

Socket sock;
MemoryStream ms = new MemoryStream();
Queue<string> strings = new Queue<string>();
// Snip init and receive connection
sock.BeginReceive(buffer, 0, MaxSize, SocketFlags.None, new AsyncCallback(OnReceived), null);

void OnReceived(IAsyncResult result) {
  // Snip sanity stuff
  int bytesReceived = sock.EndReceive(result);

  ms.Write(buffer, 0, bytesReceived);
  ms.Flush();
  long endPos = ms.Position;

  ms.Seek(0, SeekOrigin.Begin);
  long readPos = ms.Position;
  var bw = new BinaryReader(ms);

  for (;;) {
    try {
      string currentString = bw.ReadString();
      strings.Enqueue(currentString);
      readPos = stream.Position;
    }
    catch (EndOfStreamException) {
      long unusedBytes = endPos - readPos;
      var remaining = new MemoryStream();
      remaining.Write(stream.GetBuffer(), (int) readPos, (int) unusedBytes);
      ms = remaining;
      break;
    }
  }
  sock.BeginReceive(buffer, 0, MaxSize, SocketFlags.None, new AsyncCallback(OnReceived), null);
}

我相信它会正确处理拆分和合并的数据。