Question

From the internet I got the way to read a huge string from a NetworkStream.

 static NetworkStream ns = null;
 static StringBuilder sb = null;
 static byte[] buffer = null;
 static int position = 0;
 //.......................................
 //other codes skipped for simplicity
 //....................................... 
 private static string Read()
 {
        if (ns.CanRead)
        {
            sb.Clear();
            position = 0;
            while (ns.DataAvailable)
            {
                position = ns.Read(buffer, 0, buffer.Length);
                sb.Append(Encoding.Unicode.GetString(buffer, 0, position));
            }

            return sb.ToString().Trim();
        }
        else
        {
            return null;
        }
 }

However, I cannot find an example how to write a huge string to a NetworkStream.

Is there a “symmetrical” pattern for writing as we do for reading?

Thank you in advance.

Answer 1

That reading code is dangerously wrong in many ways:

• By using static variables in this way, it’s hopelessly unsuitable for multi-threaded tasks. (I hope that’s just due to you simplifying it…)
• It never initializes the variables to non-null values – again, hopefully that’s not the real code
• It uses the DataAvailable property to decide when it should be “done” – that’s incredibly dangerous as it means if a packet is delayed in the stream, you could read half as much data as you expected to
• It uses Encoding.Unicode always, which is rarely the best choice of encoding
• It assumes that it will always read a whole number of characters. What if one character is split between reads? That’s what the Encoder/Decoder classes are for… but you don’t really need to use them here anyway – see below.

I would strongly suggest that you wrap the NetworkStream in a StreamReader for reading, and a StreamWriter for writing. That’s what they’re for. You can then read a line at a time, or just a char[] buffer, or to the end of the stream (which means “until the socket is closed”). This is fine for a text-only protocol.

If you’ve got a protocol which mixes text and binary data, life becomes a lot harder. Personally I like protocols which length-prefix messages – that way you can read only the data you’re meant to, and then perform whatever conversion you want.

Anyway, I hope this random selection of thoughts helps… if you want more detailed assistance, please provide details of what protocol you’re using.