We Keep Coding

How can I return XML from BeforeSendRequest and AfterReceiveReply to the calling method in a thread-safe way?

We have a console application using the Azure WebJob SDK. The WebJob relies on a WCF service using SOAP, which it accesses through a DLL we wrote that wraps the auto-generated WCF types in something a bit more friendly.
For logging purposes, we want to save the request and response XML bodies for requests that we make. These XML bodies would be saved in our database. But, because the WCF code lives in a low-level DLL, it has no concept of our database and can't save to it.
The DLL uses Microsoft's DI extensions to register types, and the WebJob calls into it like this:
class WebJobClass
{
IWCFWrapperClient _wcfWrapperClient;
public WebJobClass(IWCFWrapperClient wcfWrapperClient)
{
_wcfWrapperClient = wcfWrapperClient;
}
public async Task DoThing()
{
var callResult = await _wcfWrapperClient.CallWCFService();
}
}
IWCFWrapperClient looks like this:
class WCFWrapperClient : IWCFWrapperClient
{
IWCF _wcf; // auto-generated by VS, stored in Reference.cs
public async Task<object> CallWCFService()
{
return await _wcf.Call(); // another auto-generated method
}
}
I've implemented an IClientMessageInspector, and it works fine to get me the XML request/response, but I don't have a way to pass it back up to WCFWrapperClient.CallWCFService so that it can be returned to WebJobClass.DoThing(), who could then save it to the database.
The problem is multithreading. WebJobs, IIRC, will run multiple requests in parallel, calling into the DLL from multiple threads. This means we can't, say, share a static property LastRequestXmlBody since multiple threads could overwrite it. We also can't, say, give each call a Guid or something since there's no way to pass anything from IWCFWrapperClient.CallWCFService into the auto-generated IWCF.Call except what was auto-generated.
So, how can I return XML to WebJobClass.DoThing in a thread-safe way?

I was able to find a solution that uses ConcurrentDictionary<TKey, TValue>, but it's a bit ugly.
First, I amended the auto-generated classes in Reference.cs with a new property Guid InternalCorrelationId. Since the auto-generated classes are partial, this can be done in separate files that aren't changed when the client is regenerated.
public partial class AutoGeneratedWCFType
{
private Guid InternalCorrelationIdField;
[System.Runtime.Serialization.DataMember()]
public Guid InternalCorrelationId
{
get { return InternalCorrelationIdField; }
set { InternalCorrelationIdField = value; }
}
}
Next, I made all my request DTO types derive from a type named RequestBase, and all my response DTO types derive from a typed named ResponseBase, so I could handle them generically:
public abstract class RequestBase
{
public Guid InternalCorrelationId { get; set; }
}
public abstract class ResponseBase
{
public string RequestXml { get; set; }
public string ResponseXml { get; set; }
}
I then added a type RequestCorrelator that simply holds on to a ConcurrentDictionary<Guid, XmlRequestResponse>:
public sealed class RequestCorrelator : IRequestCorrelator
{
public ConcurrentDictionary<Guid, XmlRequestResponse> PendingCalls { get; }
public RequestCorrelator() => PendingCalls = new ConcurrentDictionary<Guid, XmlRequestResponse>();
}
public sealed class XmlRequestResponse
{
public string RequestXml { get; set; }
public string ResponseXml { get; set; }
}
RequestCorrelator is its own type for DI purposes - you may just be able to use a ConcurrentDictionary<TKey, TValue> directly.
Finally, we have the code that actually grabs the XML, a type implementing IClientMessageInspector:
public sealed class ClientMessageProvider : IClientMessageInspector
{
private readonly IRequestCorrelator _requestCorrelator;
public ClientMessageProvider(IRequestCorrelator requestCorrelator) =>
_requestCorrelator = requestCorrelator;
public object BeforeSendRequest(ref Message request, IClientChannel channel)
{
var requestXml = request.ToString();
var internalCorrelationId = GetInternalCorrelationId(requestXml);
if (internalCorrelationId != null)
{
if (_requestCorrelator.PendingCalls.TryGetValue(internalCorrelationId.Value,
out var requestResponse))
{
requestResponse.RequestXml = requestXml;
}
request = RemoveInternalCorrelationId(request);
}
return internalCorrelationId;
}
public void AfterReceiveReply(ref Message reply, object correlationState)
{
// WCF can internally correlate a request between BeforeSendRequest and
// AfterReceiveReply. We reuse the same correlation ID we added to the
// XML as our correlation state.
var responseXml = reply.ToString();
var internalCorrelationId = (correlationState is Guid guid)
? guid
: default;
if (_requestCorrelator.PendingCalls.TryGetValue(internalCorrelationId,
out var requestResponse))
{
requestResponse.ResponseXml = responseXml;
}
}
private static Guid? GetInternalCorrelationId(string requestXml)
{
var document = XDocument.Parse(requestXml);
var internalCorrelationIdElement = /* You'll have to write this yourself;
every WCF XML request is different. */
return internalCorrelationIdElement != null
? Guid.Parse(internalCorrelationIdElement.Value)
: null;
}
private static Message RemoveInternalCorrelationId(Message oldMessage)
{
// https://stackoverflow.com/a/35639900/2709212
var buffer = oldMessage.CreateBufferedCopy(2 * 1024 * 1024);
var tempMessage = buffer.CreateMessage();
var dictionaryReader = tempMessage.GetReaderAtBodyContents();
var document = new XmlDocument();
document.Load(dictionaryReader);
dictionaryReader.Close();
var internalCorrelationIdNode = /* You'll also have to write this yourself. */
var parent = internalCorrelationIdNode.ParentNode;
parent.RemoveChild(internalCorrelationIdNode);
var memoryStream = new MemoryStream();
var xmlWriter = XmlWriter.Create(memoryStream);
document.Save(xmlWriter);
xmlWriter.Flush();
xmlWriter.Close();
memoryStream.Position = 0;
var xmlReader = XmlReader.Create(memoryStream);
var newMessage = Message.CreateMessage(oldMessage.Version, null, xmlReader);
newMessage.Headers.CopyHeadersFrom(oldMessage);
newMessage.Properties.CopyProperties(oldMessage.Properties);
return newMessage;
}
}
In short, this type:
Finds the correlation ID in the XML request.
Finds the XmlRequestResponse with the same correlation ID and adds the request to it.
Removes the correlation ID element so that the service doesn't get elements they didn't expect.
After receiving a reply, uses correlationState to find the XmlRequestResponse and write the response XML to it.
Now all we have to do is change IWCFWrapperClient:
private async Task<TDtoResult> ExecuteCallWithLogging<TDtoRequest,
TWcfRequest,
TWcfResponse,
TDtoResult>(TDtoRequest request,
Func<TDtoRequest, TWcfRequest> dtoToWcfConverter,
Func<TWcfRequest, Task<TWcfResponse>> wcfCall,
Func<TWcfResponse, TDtoResult> wcfToDtoConverter)
where TDtoRequest : CorrelationBase
where TDtoResult : WcfBase
{
request.InternalCorrelationId = Guid.NewGuid();
var xmlRequestResponse = new XmlRequestResponse();
_requestCorrelator.PendingCalls.GetOrAdd(request.InternalCorrelationId,
xmlRequestResponse);
var response = await contractingCall(dtoToWcfConverter(request));
_requestCorrelator.PendingCalls.TryRemove(request.InternalCorrelationId, out _);
return wcfToDtoConverter(response).WithRequestResponse(xmlRequestResponse);
}
public async Task<DoThingResponseDto> DoThing(DoThingRequestDto request) =>
await ExecuteCallWithLogging(request,
r => r.ToWcfModel(),
async d => await _wcf.Call(d),
d => d.ToDtoModel());
WithRequestResponse is implemented as follows:
public static T WithRequestResponse<T>(this T item, XmlRequestResponse requestResponse)
where T : ResponseBase
{
item.RequestXml = requestResponse?.RequestXml;
item.ResponseXml = requestResponse?.ResponseXml;
return item;
}
And there we go. WCF calls that return their XML in the response object rather than just something you can print to console or log to a file.

Chatbot Language Translation: Saving the Detected Language for Response Translation

I have a chatbot that is implementing translation middleware. The middleware detects the incoming language and translates the query into English. I have been working to save the detected language as a variable to pass for translating the response into the user's language, but have hit a roadblock.
protected override async Task OnMessageActivityAsync(ITurnContext<IMessageActivity> turnContext, CancellationToken cancellationToken)
{
var body = new object[] { new { Text = turnContext.Activity.Text } };
var requestBody = JsonConvert.SerializeObject(body);
//Console.WriteLine("------------------------------------------------------------------------");
//Console.WriteLine(requestBody);
//Console.WriteLine("------------------------------------------------------------------------");
//var languageChoice = "de";
using (var client = new HttpClient())
using (var request = new HttpRequestMessage())
{
//var uri = "https://api.cognitive.microsofttranslator.com" + "/translate?api-version=3.0" + "&from=" + languageChoice + "&to=" + "en";
var uri = "https://api.cognitive.microsofttranslator.com" + "/translate?api-version=3.0" + "&to=" + "en";
request.Method = HttpMethod.Post;
request.RequestUri = new Uri(uri);
request.Content = new StringContent(requestBody, Encoding.UTF8, "application/json");
request.Headers.Add("Ocp-Apim-Subscription-Key", _configuration["TranslatorKey"]);
request.Headers.Add("Ocp-Apim-Subscription-Region", "westus2");
var translatedResponse = await client.SendAsync(request);
var responseBody = await translatedResponse.Content.ReadAsStringAsync();
Console.WriteLine("------------------------------------------------------------------------");
Console.WriteLine(responseBody);
Console.WriteLine("------------------------------------------------------------------------");
var translation = JsonConvert.DeserializeObject<TranslatorResponse[]>(responseBody);
var detectedLanguage = JsonConvert.DeserializeObject<DetectLanguage[]>(responseBody);
var ourResponse = detectedLanguage?.FirstOrDefault()?.DetectedLanguage?.FirstOrDefault()?.Language.ToString();
Console.WriteLine("------------------------------------------------------------------------");
Console.WriteLine(ourResponse);
Console.WriteLine("------------------------------------------------------------------------");
//Console.WriteLine("------------------------------------------------------------------------");
//Console.WriteLine(turnContext.Activity.Text);
//Console.WriteLine(translation?.FirstOrDefault()?.Translations?.FirstOrDefault()?.Text.ToString());
//Console.WriteLine("------------------------------------------------------------------------");
// Update the translation field
turnContext.Activity.Text = translation?.FirstOrDefault()?.Translations?.FirstOrDefault()?.Text.ToString();
}
// First, we use the dispatch model to determine which cognitive service (LUIS or QnA) to use.
var recognizerResult = await _botServices.Dispatch.RecognizeAsync(turnContext, cancellationToken);
// Top intent tell us which cognitive service to use.
var topIntent = recognizerResult.GetTopScoringIntent();
// Next, we call the dispatcher with the top intent.
// ***** ERROR *****
await DispatchToTopIntentAsync(turnContext, topIntent.intent, recognizerResult, cancellationToken);
}
Using the console log from responseBody:
[{"detectedLanguage":{"language":"es","score":1.0},"translations":[{"text":"Hello","to":"en"}]}]
We have determined that responseBody is an Object array that contains an attribute “detectedLanguage”, an object, with the attributes “language” and “score”.
To retrieve the variable “language” from the object “detectedLanguage” we have made the following attempt, using “turnContext.Activity.Text” which is returned as the translated text as an example to follow.
We also added two internal classes that mimic the implementation of “turnContext.Activity.Text”:
1)
namespace Microsoft.BotBuilderSamples.Translation.Model
{
/// <summary>
/// Array of translated results from Translator API v3.
/// </summary>
internal class TranslatorResponse
{
[JsonProperty("translations")]
public IEnumerable<TranslatorResult> Translations { get; set; }
}
internal class DetectLanguage
{
[JsonProperty("detectedLanguage")]
public IEnumerable<LanguageResult> DetectedLanguage { get; set; }
}
}
using Newtonsoft.Json;
namespace Microsoft.BotBuilderSamples.Translation.Model
{
/// <summary>
/// Translation result from Translator API v3.
/// </summary>
internal class TranslatorResult
{
[JsonProperty("text")]
public string Text { get; set; }
[JsonProperty("to")]
public string To { get; set; }
}
internal class LanguageResult
{
[JsonProperty("language")]
public string Language { get; set; }
}
}
However, when we try to test and display “ourResponse” in the console we are met with the following error message:
fail: Microsoft.Bot.Builder.Integration.AspNet.Core.BotFrameworkHttpAdapter[0]
[OnTurnError] unhandled error : Cannot deserialize the current JSON object (e.g. {"name":"value"}) into type 'System.Collections.Generic.IEnumerable`1[Microsoft.BotBuilderSamples.Translation.Model.LanguageResult]' because the type requires a JSON array (e.g. [1,2,3]) to deserialize correctly.
To fix this error either change the JSON to a JSON array (e.g. [1,2,3]) or change the deserialized type so that it is a normal .NET type (e.g. not a primitive type like integer, not a collection type like an array or List<T>) that can be deserialized from a JSON object. JsonObjectAttribute can also be added to the type to force it to deserialize from a JSON object.
Path '[0].detectedLanguage.language', line 1, position 33.
We have determined that we are hitting this error because “detectedLanguage” is not an array like “translations” is. (Shown in console output below):
[{"detectedLanguage":{"language":"es","score":1.0},"translations":[{"text":"Hello","to":"en"}]}]
My question is how do we adjust this implementation to work with the object detectedLanguage outside of an array, or how do we adjust detectedLanguage to be contained within a string to work with the implementation?

Your class structure doesn't match your json. What you want is the following:
internal class DetectedLanguage
{
[JsonProperty("language")]
public string Language { get; set; }
[JsonProperty("score")]
public double Score { get; set; }
}
internal class TranslatorResult
{
[JsonProperty("text")]
public string Text { get; set; }
[JsonProperty("to")]
public string To { get; set; }
}
internal class TranslatorResponse
{
[JsonProperty("detectedLanguage")]
public DetectedLanguage DetectedLanguage { get; set; }
[JsonProperty("translations")]
public IEnumerable<TranslatorResult> Translations { get; set; }
}
Note how DetectedLanguage should be represented as a single object ON the TranslatorResponse object, which you are not doing currently. In addition, the Language property of DetectedLanguage is also a single property, not a collection.

C# Azure: How to read body from Microsoft.Azure.ServiceBus.Message?

I have trouble with the Microsoft.Azure.ServiceBus.Message class. I want to create a Message object containing a payload object and then read this object back from it. In my current example I am not even sending the Message through a real Azure bus; I'm just creating it in memory and then trying to read it.
I cannot figure out what type I am supposed to read the message body as. I've tried byte[], string and the original object type. In all my cases I get an XmlException: "The input source is not correctly formatted".
Can someone please tell me what I am doing wrong, either when encoding or decoding the Message?
[DataContract]
public class Thingy
{
[DataMember]
public string Doodad { get; set; }
}
private static Message CreateMessage()
{
var entityMessage = new Thingy {Doodad = "foobar"};
var serializedMessageBody = JsonConvert.SerializeObject(entityMessage);
var contentType = typeof(Thingy).AssemblyQualifiedName;
var bytes = Encoding.UTF8.GetBytes(serializedMessageBody);
var message = new Message(bytes) {ContentType = contentType};
return message;
}
[Test]
public void ReadMessageBytes()
{
var message = CreateMessage();
var body = message.GetBody<byte[]>();
Console.WriteLine(body);
}
[Test]
public void ReadMessageString()
{
var message = CreateMessage();
var body = message.GetBody<string>();
Console.WriteLine(body);
}
[Test]
public void ReadMessageThingy()
{
var message = CreateMessage();
var body = message.GetBody<Thingy>();
Console.WriteLine(body);
}

I found out that this works:
[Test]
public void ReadMessageProperly()
{
var message = CreateMessage();
var body = message.Body;
var text = Encoding.UTF8.GetString(body);
var thingy = JsonConvert.DeserializeObject<Thingy>(text);
Assert.IsInstanceOf<Thingy>(thingy);
Assert.AreEqual("foobar", thingy.Doodad);
}

When creating a BrokeredMessage using custom DataContract type and using DataContractSerializer :
Record recordDataContract = new Record { Id = "DataContract Record" };
BrokeredMessage recordDataContractMessage = new BrokeredMessage(recordDataContract, new DataContractSerializer(typeof(Record)));
You can receive this message as:
Record r = receiveMessage.GetBody<Record>(new DataContractSerializer(typeof(Record)));
When creating a **BrokeredMessage** using custom **DataContract** type and using default serializer (DataContract + Binary Xml):
[DataContract(Namespace = "")]
class Record {
[DataMember]
public string Id { get; set; }
}
Record recordDefault = new Record { Id = "default Record" };
BrokeredMessage recordDefaultMessage = new BrokeredMessage(recordDefault);
You can receive this message as:
Record r = receiveMessage.GetBody<Record>();
For additional reference , you can check this blog. It has detailed example for different scenarios.
Hope it helps.

Is there a construct or pattern similar to C# `using` which will return an object?

I have a WCF message inspector which inspects requests and responses: Message. The inspector works fine. A Message object can only be read once so once you read it, you cannot simply propagate as WCF will complain that the message has been read. Therefore, I am creating a brand new copy of the message and propagating that.
I have designed a class that allows message reading and after the caller has read whatever they want, they need to call Close which will return a copy of the message. Here is the skeleton of my class:
using System.ServiceModel.Channels;
internal abstract class MessageReader
{
internal string ReadSomething(string id)
{
// Return string
}
internal string ReadSomethingElse(string id)
{
// Return string
}
internal Message Close()
{
// Create copy and return it.
}
}
Users of my class may forget to call Close() which is fine because WCF will yell at them. Right now I have documentation to let users know they need to call Close().
Here is the question
Is there a pattern, or something similar, to C#'s using construct but one which returns an object at the end? This will be really convenient because then users of my class can just use a construct like that and at the end it will return the copy of the message. Something like this:
UsingSomeConstruct(var reader = new MessageReader(ref originalMessage))
{
var a = reader.ReadSomething("something");
var b = reader.ReadSomethingElse("something");
// Do something with what was read
}
// At this point originalMessage will be the copy of the message and no longer the original message.
EDIT
I thought about hacking IDisposable to achieve this but I am NOT going to do it that way so looking for other ideas.

There is no such language construct of course.
What I could suggest is to use IDisposable for cleaning up, and add ref Message message argument to each ReadXXX method. I know it will not be so convenient for your users, but from the other side they cannot forget passing the parameter.
So the implementation would be something like this:
internal class MessageReader : IDisposable
{
private MessageBuffer buffer;
private Message message;
private void Release()
{
if (buffer == null) return;
buffer.Close();
buffer = null;
message = null;
}
protected void OnReadRequest(ref Message message)
{
if (message == null) throw new ArgumentNullException("message");
if (this.message == message) return;
Release();
this.buffer = message.CreateBufferedCopy(int.MaxValue);
message = this.message = buffer.CreateMessage();
}
public void Dispose()
{
Release();
}
internal string ReadSomething(ref Message message, string id)
{
OnReadRequest(ref message);
// Return string
}
internal string ReadSomethingElse(ref Message message, string id)
{
OnReadRequest(ref message);
// Return string
}
}
and the sample usage:
using (var reader = new MessageReader())
{
var a = reader.ReadSomething(ref originalMessage, "something");
var b = reader.ReadSomethingElse(ref originalMessage, "something");
// Do something with what was read
}
// At this point originalMessage will be the copy of the message and no longer the original message.

The way I'd do this is as follows:
public MessageReader: IDisposable
{
public static MessageReader Create(ref Message message)
{
var buffer = message.CreateBufferedCopy(/*whatever is fit*/);
try
{
var reader = new MessageReader(buffer);
message = buffer.CreateMessage();
return reader;
}
catch
{
buffer.Close();
throw;
}
}
private readonly MessageBuffer buffer;
private bool disposed;
private MessageReader(MessageBuffer buffer) { this.buffer = buffer; }
public void Dispose()
{
if (disposed)
return;
buffer.Close();
disposed = true;
}
public string Read(string id)
{
var newCopy = buffer.CreateMessage();
//work with new copy...
}
}
And you'd simply use it like this:
using (var reader = MessageReader.Create(ref message))
//message here is already an untouched copy with no need of user active
//intervention and is never touched again by the reader.
{
var a = reader.Read("something"); //reads copy
...
}
IMHO, this is as clean as it can be. Note that MessageReader implements IDisposable exclusively because it holds a reference to the disposable private MessageBuffer.

Thanks to all the help from #InBetween, #quetzalcoatl, and #Ivan Stoev. Upvoted your answers because it helped me arrive at the following.
In the constructor, I create a copy of the message and set the original message to the copy. Since the status of this message is Created WCF will be happy propogating it. I create another copy and use that for reading multiple times.
#Ivan said but what if the user does not ask for anything to be read then the copying was wasted work. That is a good point but in my case, this is an interceptor and all messages are intercepted to be read.
Here is the code I ended up with suggestions from all of you:
public class MessageReader : IDisposable {
private readonly Message message;
public MessageReader(ref Message originalMessage) {
using( var buffer = originalMessage.CreateBufferedCopy( int.MaxValue ) ) {
// Keep original message for reading
this.message = buffer.CreateMessage();
// Set original message to a copy of the original
originalMessage = buffer.CreateMessage();
}
}
public int ReadSomething(string id) {
// Read from this.message;
}
public int ReadSomethingElse(string id) {
// Read from this.message;
}
public void Close() {
this.Dispose();
}
public void Dispose() {
this.message.Close();
}
}
The caller can either use it in a using block or without it. The using block is used for good reasons and not as a hack.
public object AfterReceiveRequest(ref Message request, IClientChannel channel,
InstanceContext instanceContext) {
try {
using( var rdr = new MessageReader(ref request) ) {
var value= rdr.ReadSomething( someIdentifier );
return value;
}
}
catch( System.Exception ex ) {
throw CreateFault( ex, request );
}
}

Nope, there is no such construct. It is simply too specific to exist there out of the box. There are extension methods which often are very helpful, but you won't be able to use them on this ref Message parameter..
However, if you are willing to use ref at all, then why dont simply include all that logic it in Reader's constructor?
Here's an example, somewhat contrived, but it should show what I mean. Like others mentioned in comments, I also suggest implementing IDisposable on the Reader object instead of Close, so I included that already.
TL;DR: In example below, the most important thing is in Reader(ref msg) constructor which clones the message, copies the data, and replaces the original message with a safe-message class which can be read many times..
using System;
using System.Collections.Generic;
using System.Linq;
using System.Text.RegularExpressions;
namespace Rextester
{
public class Program
{
public static void Main(string[] args)
{
// real-world variables, keep them typed as base Message
// to be able to silently replace them with different objects
Message original1;
Message original2;
// let's construct some one-time readable messages
{
var tmp1 = new OneTimeMessage();
tmp1.data["mom"] = "dad";
tmp1.data["cat"] = "dog";
original1 = tmp1;
var tmp2 = new OneTimeMessage();
tmp2.data["mom"] = "dad";
tmp2.data["cat"] = "dog";
original2 = tmp2;
}
// test1 - can't read twice
Console.WriteLine("test0A:" + original1.GetData("mom"));
//Console.WriteLine("test0B:" + original1.GetData("mom")); // fail
// test2 - can read twice with Reader's help
var backup1 = original2;
using(var rd1 = new Reader(ref original2))
{
Console.WriteLine("test1A:" + rd1.ReadSomething("mom"));
}
var backup2 = original2;
using(var rd2 = new Reader(ref original2))
{
Console.WriteLine("test1A:" + rd2.ReadSomething("mom"));
//^ ok - becase Reader replaced 'original2' with SafeMessage
}
// test3: Reader's ctor is intelligent
// so no more SafeMessages created during future usage
var backup3 = original2;
using(var rd3 = new Reader(ref original2))
{
}
var backup4 = original2;
using(var rd4 = new Reader(ref original2))
{
}
Console.WriteLine("checking for copies:" + (original2 == backup1));
Console.WriteLine("checking for copies:" + (original2 == backup2));
Console.WriteLine("checking for copies:" + (original2 == backup3));
Console.WriteLine("checking for copies:" + (original2 == backup4));
}
}
}
public abstract class Message
{
public abstract string GetData(string id);
}
public class OneTimeMessage : Message // this models your current one-time-readable message
{
public IDictionary<string, string> data = new Dictionary<string, string>();
public override string GetData(string id)
{
var tmp = data[id];
data.Remove(id);
// that's nonsense, but I want to show that you can't
// read the same thing twice from this object
return tmp;
}
}
public class SafeMessage : Message
{
public IDictionary<string, string> data;
public override String GetData(string id)
{
return data[id];
}
public SafeMessage(Message msg)
{
// read out the full msg's data and store it
// since this is example, we can do it in a pretty simple way
// in your code that will probably be more complex
this.data = new Dictionary<string,string>(((OneTimeMessage)msg).data);
}
}
public class Reader : IDisposable
{
private Message message;
public Reader(ref Message src)
{
src = src is SafeMessage ? src : new SafeMessage(src);
this.message = src;
}
public string ReadSomething(string id){ return message.GetData(id); }
public void Dispose(){ Close(); }
public void Close(){ message=null; Console.WriteLine("reader closed"); }
}
EDIT: improved example
using System;
using System.Collections.Generic;
using System.Linq;
using System.ServiceModel.Channels;
using System.Text.RegularExpressions;
using System.Xml;
namespace MyProgram
{
public class Program
{
public static void Main(string[] args)
{
// real-world variables, keep them typed as base Message
// to be able to silently replace them with different objects
Message original1;
Message original2;
// let's construct some one-time readable messages
{
original1 = new TheMessage("dad", "dog");
original2 = new TheMessage("dad", "dog");
}
// test1 - can't read twice
Console.WriteLine("test0A:" + original1.GetReaderAtBodyContents().ReadOuterXml());
// Console.WriteLine("test0B:" + original1.GetReaderAtBodyContents().ReadOuterXml()); // fail: InvalidOperationException - it was already read
// test2 - can read ONCE with Reader's help, but the message is replaced and is usable again
var backup1 = original2;
using (var rd1 = new ReaderOnce(ref original2))
{
Console.WriteLine("is message replaced after opening Reader:" + (original2 != backup1));
Console.WriteLine("test1A:" + rd1.ReadBodyXml());
// Console.WriteLine("test1B:" + rd1.ReadBodyXml()); // fail: InvalidOperationException - it was already read
}
// test3 - can read MANY TIMES with ReaderMany's help
// also note we use 'original2' again, which was already used above, so in fact ReaderOnce really works as well
var backup2 = original2;
using (var rd1 = new ReaderMany(ref original2))
{
Console.WriteLine("is message replaced after opening Reader:" + (original2 != backup2));
Console.WriteLine("test2A:" + rd1.ReadBodyXml());
Console.WriteLine("test2B:" + rd1.ReadBodyXml()); // ok
}
Console.WriteLine("Press enter to exit");
Console.ReadLine();
}
}
}
// solution1
public class ReaderOnce : IDisposable
{
private Message localCopy;
public ReaderOnce(ref Message src)
{
// create a WCF MessageBuffer to assist in copying messages
// btw. I suppose you should set some sane limit instead of that below
using (var tempBuffer = src.CreateBufferedCopy(int.MaxValue))
{
src = tempBuffer.CreateMessage(); // FIRST copy for outer use
localCopy = tempBuffer.CreateMessage(); // SECOND copy for internal use in the Reader
}
}
public void Dispose() { Close(); }
public void Close()
{
localCopy.Close(); // but that does NOT affect FIRST copy sent to outer scope outside reader
Console.WriteLine("reader closed");
}
public string ReadBodyXml() // careful: that's again ONE TIME readable
{
return localCopy.GetReaderAtBodyContents().ReadOuterXml();
}
}
// solution2
public class ReaderMany : IDisposable
{
private MessageBuffer localBuffer;
public ReaderMany(ref Message src)
{
localBuffer = src.CreateBufferedCopy(int.MaxValue);
src = localBuffer.CreateMessage(); // FIRST copy for outer use
}
public void Dispose() { Close(); }
public void Close()
{
localBuffer.Close();
Console.WriteLine("reader closed");
}
public string ReadBodyXml() // this is readable multiple times
{
using (var tmp = localBuffer.CreateMessage())
return tmp.GetReaderAtBodyContents().ReadOuterXml();
}
}
// let's fake some Message type to have something to test the Reader on
public class TheMessage : Message
{
public override MessageHeaders Headers => _mh;
public override MessageProperties Properties => _mp;
public override MessageVersion Version => _mv;
private MessageHeaders _mh;
private MessageProperties _mp;
private MessageVersion _mv;
private string data1;
private string data2;
// btw. below: surprise! XmlDictionaryWriter is in "System.Runtime.Serialization", not in "System.Xml"
protected override void OnWriteBodyContents(XmlDictionaryWriter writer)
{
writer.WriteStartElement("foo");
writer.WriteAttributeString("data1", data1);
writer.WriteAttributeString("data2", data2);
writer.WriteEndElement();
}
public TheMessage(string data1, string data2)
{
// remember, this class is just an example, you will work on your own messages you already have
_mv = MessageVersion.Soap12;
_mh = new MessageHeaders(_mv);
_mp = new MessageProperties();
// below: yeah, that's super-naive and wrong, but that's an example
this.data1 = data1;
this.data2 = data2;
}
}

There is no language construct in c# that does what you are asking. As stated in comments, you could abuse IDisposable and the language and use a using block to achieve what you want.
But, I fail see what you are gaining, you are just punting the problem; now users will need to remember to use usinginstead of Close. The latter is simple and clean, the former uses a very known language construct to do something different to what it was thought for, something that will potentially be very confusing.

Best way to implement a request response pattern that is extendable

OK,
I have to create a C# library that can send commands to a device and process the command specific responses and broadcasts over a serial port (or other communications method). The library must also be able to handle request and response extensions held in other libraries as certain devices implement an extended command set, but it must be possible to choose whether these extended commands are utilised (I guess using reflection in the client app). I have created a class of type Packet that is able to create the packet and add its payload, calculate its checksum and write the packet to the stream.
public class Packet
{
internal PacketHeaderType Header { get; private set; }
internal List<byte> Payload { get; private set; }
protected int PayloadLength { get { return Payload.Count; } }
protected byte HeaderByte { get { return (byte)((Convert.ToByte(Header) << 4) | PayloadLength); } } //we need to add the packet length to the lower nibble of the header before sending
public Packet(PacketHeaderType header, List<byte> payload)
{
this.Header = header;
this.Payload = new List<byte>(payload);
}
public Packet(PacketHeaderType headerByte)
{
this.Header = headerByte;
this.Payload = new List<byte>();
}
internal byte XorByte
{
get
{
Byte xorByte = Convert.ToByte(HeaderByte);
for (int i = 0; i < PayloadLength; i++)
xorByte ^= Payload.ToArray()[i];
return xorByte;
}
}
public async Task WriteAsync(Stream stream, bool flush = true, CancellationToken token = default(CancellationToken))
{
var buffer = new List<byte>();
buffer.Add(HeaderByte);
if (Payload != null && PayloadLength > 0)
{
buffer.AddRange(Payload);
}
buffer.Add(XorByte);
await stream.WriteAsync(buffer.ToArray(), 0, buffer.Count);
if (flush)
{
await stream.FlushAsync();
}
}
}
I have also created child classes that implement Type packet for each of the valid commands.
Finally I have also created a class of type PacketHandler that is able to read bytes from a stream and create it into a packet object.
The way I envisage using the library would be like this:
public async string GetCmdStnSoftwareVersion()
{
var msgReq = new CmdStnSoftwareVersionReqMessage();
await msgReq.WriteAsync(sPort.BaseStream);
await var response = msgReq.GetResponse(5); //5 = timeout in seconds!
return String.Format("{0}.{1}", response.Major, response.Minor);
}
What I am stuck on is a good pattern and/or example implementation for handling responses which is compatible with implementing the extension libraries. Can anyone provide input?