Writing to then reading from a MemoryStream

Question

I'm using DataContractJsonSerializer , which likes to output to a Stream. I want to top-and-tail the outputs of the serializer so I was using a StreamWriter to alternately write in the extra bits I needed.

var ser = new DataContractJsonSerializer(typeof (TValue));

using (var stream = new MemoryStream())
{   
    using (var sw = new StreamWriter(stream))
    {
        sw.Write("{");

        foreach (var kvp in keysAndValues)
        {
            sw.Write("'{0}':", kvp.Key);
            ser.WriteObject(stream, kvp.Value);
        }

        sw.Write("}");
    }

    using (var streamReader = new StreamReader(stream))
    {
        return streamReader.ReadToEnd();
    }
}

When I do this I get an ArgumentException "Stream was not readable".

I'm probably doing all sorts wrong here so all answers welcome. Thanks.

Answer 1

Three things:

• Don't close the StreamWriter . That will close the MemoryStream . You do need to flush the writer though.
• Reset the position of the stream before reading.
• If you're going to write directly to the stream, you need to flush the writer first.

So:

using (var stream = new MemoryStream())
{
    var sw = new StreamWriter(stream);
    sw.Write("{");

    foreach (var kvp in keysAndValues)
    {
        sw.Write("'{0}':", kvp.Key);
        sw.Flush();
        ser.WriteObject(stream, kvp.Value);
    }    
    sw.Write("}");            
    sw.Flush();
    stream.Position = 0;

    using (var streamReader = new StreamReader(stream))
    {
        return streamReader.ReadToEnd();
    }
}

There's another simpler alternative though. All you're doing with the stream when reading is converting it into a string. You can do that more simply:

return Encoding.UTF8.GetString(stream.GetBuffer(), 0, (int) stream.Length);

Unfortunately MemoryStream.Length will throw if the stream has been closed, so you'd probably want to call the StreamWriter constructor that doesn't close the underlying stream, or just don't close the StreamWriter .

I'm concerned by you writing directly to the the stream - what is ser ? Is it an XML serializer, or a binary one? If it's binary, your model is somewhat flawed - you shouldn't mix binary and text data without being very careful about it. If it's XML, you may find that you end up with byte-order marks in the middle of your string, which could be problematic.

Answer 2

setting the memory streams position to the beginning might help.

 stream.Position = 0; 

But the core problem is that the StreamWriter is closing your memory stream when it is closed.

Simply flushing that stream where you end the using block for it and only disposing of it fter you have read the data out of the memory stream will solve this for you.

You may also want to consider using a StringWriter instead...

using (var writer = new StringWriter())
{
    using (var sw = new StreamWriter(stream))
    {
        sw.Write("{");

        foreach (var kvp in keysAndValues)
        {
            sw.Write("'{0}':", kvp.Key);
            ser.WriteObject(writer, kvp.Value);
        }
        sw.Write("}");
    }

    return writer.ToString();
}

This would require your serialization WriteObject call can accept a TextWriter instead of a Stream.

Answer 3

To access the content of a MemoryStream after it has been closed use the ToArray() or GetBuffer() methods. The following code demonstrates how to get the content of the memory buffer as a UTF8 encoded string.

byte[] buff = stream.ToArray(); 
return Encoding.UTF8.GetString(buff,0,buff.Length);

Note: ToArray() is simpler to use than GetBuffer() because ToArray() returns the exact length of the stream, rather than the buffer size (which might be larger than the stream content). ToArray() makes a copy of the bytes.

Note: GetBuffer() is more performant than ToArray() , as it doesn't make a copy of the bytes. You do need to take care about possible undefined trailing bytes at the end of the buffer by considering the stream length rather than the buffer size. Using GetBuffer() is strongly advised if stream size is larger than 80000 bytes because the ToArray copy would be allocated on the Large Object Heap where it's lifetime can become problematic.

It is also possible to clone the original MemoryStream as follows, to facilitate accessing it via a StreamReader eg

using (MemoryStream readStream = new MemoryStream(stream.ToArray()))
{
...
}

The ideal solution is to access the original MemoryStream before it has been closed, if possible.

Answer 4

Just a wild guess: maybe you need to flush the streamwriter? Possibly the system sees that there are writes "pending". By flushing you know for sure that the stream contains all written characters and is readable.