Develop Reference

RX - rethrow an error in containing method

I need to translate an error in an RX stream (IObservable) into an exception in the method that contains the subscription to the stream
(because of this issue https://github.com/aspnet/SignalR/pull/1331 , Whereby errors arent serialised to clients.) Once this issue is fixed I will revert to handling error properly
e.g.
I have the following method
public IObservable<StreamItem> LiveStream()
{
_mySvc.Start();
return _mySvc.ThingChanged();
}
So I have tried to subscribe to the stream and rethrow the error, but it still doesnt get transmitted to the client:
public IObservable<StreamItem> LiveStream()
{
_mySvc.Start();
_mySvc.ThingChanged().Subscribe(item => {}, OnError, () => {});
return _mySvc.ThingChanged();
}
private void OnError(Exception exception)
{
throw new Exception(exception.Message);
}
What I need is the equivelent of throwing in the LiveStream method
e.g. this error is propogated to the client
public IObservable<StreamItem> LiveStream()
{
_mySvc.Start();
throw new Exception("some error message");
return _mySvc.ThingChanged();
}
any ideas how to achieve this?

I have found this as well, especially with a "contained" reactive pipeline—that is, one with a well-defined beginning and end. In situations like those, it may suffice to simply allow underlying exceptions to bubble up to the containing scope. But as you have found, that concept is rather foreign to Rx generally: what happens in the pipeline stays in the pipeline.
The only way out of this that I have found in a contained scenario is to "slip" the error out of the stream using Catch(), and hand back an empty IObservable to allow the stream to halt naturally (otherwise, you'll hang if you're awaiting an IObservable for completion).
This will not work within your LiveStream() method, because that context/scope should have passed out of existence long before you're consuming your stream. So, this will have to happen in the context that contains the whole pipeline.
Exception error = null;
var source = LiveStream()
.Catch<WhatYoureStreaming, Exception>(ex => {error = ex; return Observable.Empty<WhatYoureStreaming>(); })
...
await source; // if this is how you're awaiting completion
// not a real exception type, use your own
if (error != null) throw new ContainingException("oops", error);
Just don't throw error there at the end, you'll lose the original stack trace.

Try this code:
public IObservable<StreamItem> LiveStream()
{
_mySvc.Start();
return
_mySvc
.ThingChanged()
.Materialize()
.Do(x =>
{
if (x.Kind == NotificationKind.OnError)
{
OnError(x.Exception);
}
})
.Dematerialize();
}
I'm not sure that this is the best way to go - throwing exceptions like this can cause you grief inside a stream where you end up with the wrong exception handlers firing. You might need to find another solution.

"yield return" from event handler

I have a class which takes a stream in the constructor. You can then set up callbacks for various events, and then call StartProcessing. The issue is that I want to use it from a function which should return an IEnumerable.
Example:
public class Parser
{
public Parser(System.IO.Stream s) { // saves stream and does some set up }
public delegate void OnParsedHandler(List<string> token);
public event OnParsedHandler OnParsedData;
public void StartProcessing()
{
// reads stream and makes callback when it has a whole record
}
}
public class Application
{
public IEnumerable<Thing> GetThings(System.IO.Stream s)
{
Parser p = new Parser(s);
p.OnParsedData += (List<string> str) =>
{
Thing t = new Thing(str[0]);
// here is where I would like to yield
// but I can't
yield return t;
};
p.StartProcessing();
}
}
Right now my solution, which isn't so great, is to put them all the Things into a List which is captured by the lambda, and then iterate over them after calling StartProcessing.
public class Application
{
public IEnumerable<Thing> GetThings(System.IO.Stream s)
{
Parser p = new Parser(s);
List<Thing> thingList = new List<Thing>();
p.OnParsedData += (List<string> str) =>
{
Thing t = new Thing(str[0]);
thingList .Add(t);
};
p.StartProcessing();
foreach(Thing t in thingList )
{
yield return t;
}
}
}
The issue here is that now I have to save all of the Thing objects into list.

The problem you have here is that you don't fundamentally have a "pull" mechanic here, you're trying to push data from the parser. If the parser is going to push data to you, rather than letting the caller pull the data, then GetThings should return an IObservable, rather than an IEnumerable, so the caller can consume the data when it's ready.
If it really is important to have a pull mechanic here then Parser shouldn't fire an event to indicate that it has new data, but rather the caller should be able to ask it for new data and have it get it; it should either return all of the parsed data, or itself return an IEnumerable.

Interesting question. I would like to build upon what #servy has said regarding push and pull. In your implementation above, you are effectively adapting a push mechanism to a pull interface.
Now, first things first. You have not specified whether the call to the StartProcessing() method is a blocking call or not. A couple of remarks regarding that:
If the method is blocking (synchronous), then there is really no point in adapting it to a pull model anyway. The caller will see all the data processed in a single blocking call.
In that regard, receiving the data indirectly via an event handler scatters into two seemingly unrelated constructs what should otherwise be a single, cohesive, explicit operation. For example:
void ProcessAll(Action<Thing> callback);
On the other hand, if the StartProcessing() method actually spawns a new thread (maybe better named BeginProcessing() and follow the Event-based Asynchronous Pattern or another async processing pattern), you could adapt it to a pull machanism by means of a synchronization construct using a wait handle: ManualResetEvent, mutex and the like. Pseudo-code:
public IEnumerable<Thing> GetThings(System.IO.Stream s)
{
var parser = new Parser(s);
var waitable = new AutoResetEvent(false);
Thing item = null;
parser.OnParsedData += (Thing thing) =>
{
item = thing;
waitable.Set();
};
IAsyncResult result = parser.BeginProcessing();
while (!result.IsCompleted)
{
waitable.WaitOne();
yield return item;
}
}
Disclaimer
The above code serves only as a means for presenting an idea. It is not thread-safe and the synchronization mechanics do not work properly. See the producer-consumer pattern for more information.

How to manage observable subscription for dependent observables?

This sample console application has 2 observables. The first one pushes numbers from 1 to 100. This observable is subscribed by the AsyncClass which runs a long running process for each number it gets. Upon completion of this new async process I want to be able to 'push' to 2 subscribers which would be doing something with this new value.
My attempts are commented in the source code below.
AsyncClass:
class AsyncClass
{
private readonly IConnectableObservable<int> _source;
private readonly IDisposable _sourceDisposeObj;
public IObservable<string> _asyncOpObservable;
public AsyncClass(IConnectableObservable<int> source)
{
_source = source;
_sourceDisposeObj = _source.Subscribe(
ProcessArguments,
ExceptionHandler,
Completed
);
_source.Connect();
}
private void Completed()
{
Console.WriteLine("Completed");
Console.ReadKey();
}
private void ExceptionHandler(Exception exp)
{
throw exp;
}
private void ProcessArguments(int evtArgs)
{
Console.WriteLine("Argument being processed with value: " + evtArgs);
//_asyncOpObservable = LongRunningOperationAsync("hello").Publish();
// not going to work either since this creates a new observable for each value from main observer
}
// http://rxwiki.wikidot.com/101samples
public IObservable<string> LongRunningOperationAsync(string param)
{
// should not be creating an observable here, rather 'pushing' values?
return Observable.Create<string>(
o => Observable.ToAsync<string, string>(DoLongRunningOperation)(param).Subscribe(o)
);
}
private string DoLongRunningOperation(string arg)
{
return "Hello";
}
}
Main:
static void Main(string[] args)
{
var source = Observable
.Range(1, 100)
.Publish();
var asyncObj = new AsyncClass(source);
var _asyncTaskSource = asyncObj._asyncOpObservable;
var ui1 = new UI1(_asyncTaskSource);
var ui2 = new UI2(_asyncTaskSource);
}
UI1 (and UI2, they're basically the same):
class UI1
{
private IConnectableObservable<string> _asyncTaskSource;
private IDisposable _taskSourceDisposable;
public UI1(IConnectableObservable<string> asyncTaskSource)
{
_asyncTaskSource = asyncTaskSource;
_asyncTaskSource.Connect();
_taskSourceDisposable = _asyncTaskSource.Subscribe(RefreshUI, HandleException, Completed);
}
private void Completed()
{
Console.WriteLine("UI1: Stream completed");
}
private void HandleException(Exception obj)
{
Console.WriteLine("Exception! "+obj.Message);
}
private void RefreshUI(string obj)
{
Console.WriteLine("UI1: UI refreshing with value "+obj);
}
}
This is my first project with Rx so let me know if I should be thinking differently. Any help would be highly appreciated!

I'm going to let you know you should be thinking differently... :) Flippancy aside, this looks like a case of bad collision between object-oriented and functional-reactive styles.
It's not clear what the requirements are around timing of the data flow and caching of results here - the use of Publish and IConnectableObservable is a little confused. I'm going to guess you want to avoid the 2 downstream subscriptions causing the processing of a value being duplicated? I'm basing some of my answer on that premise. The use of Publish() can achieve this by allowing multiple subscribers to share a subscription to a single source.
Idiomatic Rx wants you to try and keep to a functional style. In order to do this, you want to present the long running work as a function. So let's say, instead of trying to wire your AsyncClass logic directly into the Rx chain as a class, you could present it as a function like this contrived example:
async Task<int> ProcessArgument(int argument)
{
// perform your lengthy calculation - maybe in an OO style,
// maybe creating class instances and invoking methods etc.
await Task.Delay(TimeSpan.FromSeconds(1));
return argument + 1;
}
Now, you can construct a complete Rx observable chain calling this function, and through the use of Publish().RefCount() you can avoid multiple subscribers causing duplicate effort. Note how this separates concerns too - the code processing the value is simpler because the reuse is handled elsewhere.
var query = source.SelectMany(x => ProcessArgument(x).ToObservable())
.Publish().RefCount();
By creating a single chain for subscribers, the work is only started when necessary on subscription. I've used Publish().RefCount() - but if you want to ensure values aren't missed by the second and subsequent subscribers, you could use Replay (easy) or use Publish() and then Connect - but you'll want the Connect logic outside the individual subscriber's code because you just need to call it once when all subscribers have subscribed.

Writing code to fire the last method to throw an exception in a multi-threaded web app

I was writing some try-catch blocks for various methods today, and thought to myself it would be good to have utility method which would automatically call the method again for a number of times specified in a parameter, at a certain time.
However, I thought to myself, the method/property etc which will cause an exception will be at the top of the stacktrace (do property calls get put on the stacktrace?) in a single threaded application (so an application with no code relating to threading). So I can simply get the method name at the top and dynamically call it again.
So I would have code like:
string s = StackTrace.GetFrame(0).GetMethodName; (I can't remember the exact syntax).
With this method, I can execute it using an activator or one of several other ways.
But in a multi-threaded application, I could have several methods firing at once and I wouldn't know which one finishes first/last. So I can't expect a method for which I write a try-catch block to be at the top of the stack.
How would I go about achieving this?

Please don't do this. It's a really, really, really, really, really bad idea.
Maybe not as bad as deleting files randomly, if the hard drive runs out of room - but just about as bad.

While I question the need for an auto retrying mechanism (does randomly retrying really help you out in so many situations that you need a utility method?) - using StackTrace and Reflection is, at best, a terribly complicated solution.
Not that I suggest that anyone actually use this code, but I'd probably go with a delegate based approach to this particular problem:
public static class Extensions {
public static void Try(this Action a, int maxTries) {
new (Func<bool>(() => { a(); return true; })).Try(maxTries);
}
public static TResult Try<TResult>(this Func<TResult> f, int maxTries) {
Exception lastException = null;
for (int i = 0; i < maxTries; i++) {
try {
return f();
} catch (Exception ex) {
lastException = ex;
}
}
throw lastException;
}
}
Usage is a bit unorthodox, but fairly clear I think:
// Set a property
new Action(() => myObject.Property = 5).Try(5);
// With a return value
var count = new Func<int>(() => myList.Count).Try(3);
You can't inline a lambda to a method, but you could have a somewhat fluent interface:
Utilities.Try(
() => MyObject.Property = 5
).Repeat(5);
And multi line methods:
Utilities.Try(() => {
MyObject.Property1 = 5;
MyObject.Property2 = 6;
MyObject.Property3 = 7;
}).Repeat(5);

Mark's code is probably better, but here's mine...
If you really want to do something like this, I'd use code something like this. Yes, you still have to manually call it, but your idea of indiscriminately retrying ALL excepting methods is a really, really bad idea.
public class TryAgain
{
public delegate void CodeToTryAgain ();
public static void Repeat<E>(int count, CodeToTryAgain code) where E : Exception
{
while (count-- > 0)
{
try
{
code();
return;
}
catch (E ex)
{
Console.WriteLine("Caught an {0} : {1}", typeof(E).Name, ex.Message);
// ignoring it!
}
}
}
}
And then you'd call your failing method, ThrowTwice, or whatever you want to do, like this:
TryAgain.Repeat<MyException>(5, delegate()
{
ThrowTwice();
});
In this example, the Repeat method will ignore all exceptions of type MyException, trying to call ThrowTwice up to 5 times...
You can add your own sleeping and time-outs, and whatever.

Reducing duplicate error handling code in C#?

I've never been completely happy with the way exception handling works, there's a lot exceptions and try/catch brings to the table (stack unwinding, etc.), but it seems to break a lot of the OO model in the process.
Anyway, here's the problem:
Let's say you have some class which wraps or includes networked file IO operations (e.g. reading and writing to some file at some particular UNC path somewhere). For various reasons you don't want those IO operations to fail, so if you detect that they fail you retry them and you keep retrying them until they succeed or you reach a timeout. I already have a convenient RetryTimer class which I can instantiate and use to sleep the current thread between retries and determine when the timeout period has elapsed, etc.
The problem is that you have a bunch of IO operations in several methods of this class, and you need to wrap each of them in try-catch / retry logic.
Here's an example code snippet:
RetryTimer fileIORetryTimer = new RetryTimer(TimeSpan.FromHours(10));
bool success = false;
while (!success)
{
try
{
// do some file IO which may succeed or fail
success = true;
}
catch (IOException e)
{
if (fileIORetryTimer.HasExceededRetryTimeout)
{
throw e;
}
fileIORetryTimer.SleepUntilNextRetry();
}
}
So, how do you avoid duplicating most of this code for every file IO operation throughout the class? My solution was to use anonymous delegate blocks and a single method in the class which executed the delegate block passed to it. This allowed me to do things like this in other methods:
this.RetryFileIO( delegate()
{
// some code block
} );
I like this somewhat, but it leaves a lot to be desired. I'd like to hear how other people would solve this sort of problem.

This looks like an excellent opportunity to have a look at Aspect Oriented Programming. Here is a good article on AOP in .NET. The general idea is that you'd extract the cross-functional concern (i.e. Retry for x hours) into a separate class and then you'd annotate any methods that need to modify their behaviour in that way. Here's how it might look (with a nice extension method on Int32)
[RetryFor( 10.Hours() )]
public void DeleteArchive()
{
//.. code to just delete the archive
}

Just wondering, what do you feel your method leaves to be desired? You could replace the anonymous delegate with a.. named? delegate, something like
public delegate void IoOperation(params string[] parameters);
public void FileDeleteOperation(params string[] fileName)
{
File.Delete(fileName[0]);
}
public void FileCopyOperation(params string[] fileNames)
{
File.Copy(fileNames[0], fileNames[1]);
}
public void RetryFileIO(IoOperation operation, params string[] parameters)
{
RetryTimer fileIORetryTimer = new RetryTimer(TimeSpan.FromHours(10));
bool success = false;
while (!success)
{
try
{
operation(parameters);
success = true;
}
catch (IOException e)
{
if (fileIORetryTimer.HasExceededRetryTimeout)
{
throw;
}
fileIORetryTimer.SleepUntilNextRetry();
}
}
}
public void Foo()
{
this.RetryFileIO(FileDeleteOperation, "L:\file.to.delete" );
this.RetryFileIO(FileCopyOperation, "L:\file.to.copy.source", "L:\file.to.copy.destination" );
}

You could also use a more OO approach:
Create a base class that does the error handling and calls an abstract method to perform the concrete work. (Template Method pattern)
Create concrete classes for each operation.
This has the advantage of naming each type of operation you perform and gives you a Command pattern - operations have been represented as objects.

Here's what I did recently. It has probably been done elsewhere better, but it seems pretty clean and reusable.
I have a utility method that looks like this:
public delegate void WorkMethod();
static public void DoAndRetry(WorkMethod wm, int maxRetries)
{
int curRetries = 0;
do
{
try
{
wm.Invoke();
return;
}
catch (Exception e)
{
curRetries++;
if (curRetries > maxRetries)
{
throw new Exception("Maximum retries reached", e);
}
}
} while (true);
}
Then in my application, I use c#'s Lamda expression syntax to keep things tidy:
Utility.DoAndRetry( () => ie.GoTo(url), 5);
This calls my method and retries up to 5 times. At the fifth attempt, the original exception is rethrown inside of a retry exception.