Rewind request body stream

问题:

I am re-implementing a request logger as Owin Middleware which logs the request url and body of all incoming requests. I am able to read the body, but if I do the body parameter in my controller is null.

I'm guessing it's null because the stream position is at the end so there is nothing left to read when it tries to deserialize the body. I had a similar issue in a previous version of Web API but was able to set the Stream position back to 0. This particular stream throws a This stream does not support seek operations exception.

In the most recent version of Web API 2.0 I could call Request.HttpContent.ReadAsStringAsync()inside my request logger, and the body would still arrive to the controller in tact.

How can I rewind the stream after reading it?

or

How can I read the request body without consuming it?

public class RequestLoggerMiddleware : OwinMiddleware
{
    public RequestLoggerMiddleware(OwinMiddleware next)
        : base(next)
    {
    }

    public override Task Invoke(IOwinContext context)
    {
        return Task.Run(() => {
            string body = new StreamReader(context.Request.Body).ReadToEnd();
            // log body

            context.Request.Body.Position = 0; // cannot set stream position back to 0
            Console.WriteLine(context.Request.Body.CanSeek); // prints false
            this.Next.Invoke(context);
        });
    }
}

---

public class SampleController : ApiController 
{
    public void Post(ModelClass body)
    {
        // body is now null if the middleware reads it
    }
}

回答1:

Just found one solution. Replacing the original stream with a new stream containing the data.

    public override Task Invoke(IOwinContext context)
    {
        return Task.Run(() => {
            string body = new StreamReader(context.Request.Body).ReadToEnd();
            // log body

            byte[] requestData = Encoding.UTF8.GetBytes(body);
            context.Request.Body = new MemoryStream(requestData);
            this.Next.Invoke(context);
        });
    }

If you are dealing with larger amounts of data, I'm sure a FileStream would also work as the replacement.

回答2:

Here's a small improvement to the first answer by Despertar, which helped me a lot, but I ran into an issue when working with binary data. The intermediate step of extracting the stream into a string and then putting it back into a byte array using Encoding.UTF8.GetBytes(body) messes up the binary content (the contents will change unless it is an UTF8 encoded string). Here's my fix using Stream.CopyTo():

    public override async Task Invoke(IOwinContext context)
    {
        // read out body (wait for all bytes)
        using (var streamCopy = new MemoryStream())
        {
            context.Request.Body.CopyTo(streamCopy);
            streamCopy.Position = 0; // rewind

            string body = new StreamReader(streamCopy).ReadToEnd();
            // log body

            streamCopy.Position = 0; // rewind again
            context.Request.Body = streamCopy; // put back in place for downstream handlers

            await this.Next.Invoke(context);
        }
    }

Also, MemoryStream is nice because you can check the stream's length before logging the full body (which is something I don't want to do in case someone uploads a huge file).

回答3:

I know this is old, but just to help anyone that comes across this. You need to seek on the stream: context.Request.Body.Seek(0, SeekOrigin.Begin);