Let's say I have three products in a list. In order to enable a certain action, all three need to be of a certain type. In order to find out the type of the product, I need to make a service call and wait for a response.
What I would like to do is wait for all three responses (maybe with a timeout in case something goes wrong) and when all the info is gathered, decide whether or not to enable the possible action.
I used to solve this by having some counter or reset events to keep track of the finished events but I would like to see if I can use Rx to do it in a cleaner way.
As I am not too familiar with Rx yet, I am looking for some tips/pointers. I understand I can use
Observable.FromEventPattern
for the events I am waiting on. I subscribe and wait for the response and handle it. I am just not clear on how to combine the multiple events.
The combinator you are looking for is CombineLatest
Say you've got a class like this:
public class Foo
{
public delegate void FooEventHandler(object sender, EventArgs args);
public event FooEventHandler FirstEvent = delegate {};
public event FooEventHandler SecondEvent = delegate {};
public event FooEventHandler ThirdEvent = delegate {};
public void DoIt()
{
FireOne();
FireTwo();
FireThree();
}
public void FireOne()
{
Console.WriteLine("Firing event 1...");
Thread.Sleep(1000);
FirstEvent(this, new EventArgs());
}
public void FireTwo()
{
Console.WriteLine("Firing event 2...");
Thread.Sleep(1000);
SecondEvent(this, new EventArgs());
}
public void FireThree()
{
Console.WriteLine("Firing event 3...");
Thread.Sleep(1000);
ThirdEvent(this, new EventArgs());
}
}
First you'd want to "convert" those events to Observable:
var foo = new Foo();
var firstWatcher = Observable.FromEventPattern(foo, "FirstEvent");
var secondWatcher = Observable.FromEventPattern(foo, "SecondEvent");
var thirdWatcher = Observable.FromEventPattern(foo, "ThirdEvent");
Now you'll want the "Only fire when all these have fired" selector, which is CombineLatest:
var allDone = Observable.CombineLatest(firstWatcher, secondWatcher, thirdWatcher);
And to test it out:
using(allDone.Subscribe(_ => Console.WriteLine("Boop! You sunk my battleship!")))
{
foo.DoIt();
}
Alternative "test harness":
var foo = new Foo();
var firstWatcher = Observable.FromEventPattern(foo, "FirstEvent");
var secondWatcher = Observable.FromEventPattern(foo, "SecondEvent");
var thirdWatcher = Observable.FromEventPattern(foo, "ThirdEvent");
var allDone = Observable.CombineLatest(firstWatcher, secondWatcher, thirdWatcher);
// keep a handle on the subscription
IDisposable subscription = null;
// to prevent premature exiting...
var blocker = new ManualResetEvent(false);
// explicit subscribe
subscription = allDone.Subscribe(
whoCares =>
{
Console.WriteLine("BOOM! We're done!");
// always clean up after yourself
if(subscription != null)
{
subscription.Dispose();
}
// it's ok, we can quit now
blocker.Set();
});
foo.DoIt();
// Wait until it's clear to go ahead...
blocker.WaitOne();
Related
What is the best solution to quickly cancel long running processes inside background worker?
For example, we have such situation:
private void DoWork(object sender, DoWorkEventArgs e)
{
...
for (int i = 0; i < items; i++)
{
if (_worker.CancellationPending == true)
{
e.Cancel = true;
break;
}
else
{
VeryLongRunningProcess();
}
}
}
private void VeryLongRunningProcess()
{
var a = Test();
var b = Test2();
Thread.Sleep(5000);
var c = Test3();
}
In such case, VeryLongRunningProcess() will be not finished on pressing cancel until he finished everything inside his body.
What to do in such cases?
I tried to pass (BackgroundWorker)sender to VeryLongRunningProcess() as param and inside this method check for CancellationPending, but i dont know is this correct way or not
If the problem is isolated your VeryLongRunningProcess from classes like the worker, you can use a Func as a parameter and leave outside your method the worker access
private void VeryLongRunningProcess(Func<bool> isCancelled)
{
var a = Test();
if (isCancelled())
{
return;
}
var b = Test2();
if (isCancelled())
{
return;
}
Thread.Sleep(5000);
var c = Test3();
}
Inside your method, you may check if you must cancel the operation as many times you need. And you can use the Func as a parameter in other methods like Test1, Test2... if any of them takes long time to finish.
Then, you invoke your method in this form:
VeryLongRunningProcess(() => _worker.CancellationPending);
As other people comment, maybe interesting use async/await.
UPDATE
Another way to do if you want choose the use or not of the cancellation:
private void VeryLongRunningProcess(Func<bool> isCancelled = null)
{
var a = Test();
// Or: isCancelled != null && isCancelled()
if (isCancelled?.Invoke() ?? false)
{
return;
}
// ...
}
Normally you should create long-running process as "async" method (public async Task or Task DoWork()) for resources destribution purposes. "CancelationToken" enables cooperative cancellation between threads, thread pool work items. Also it is possible to propagate a callback delegate that can be invoked when Cancellation Token cancelled or function is compleete.
I have a service that raises multiple events, some of them can be raised at the same time. I need to process those events and run a potentially long running method based on the event arguments.
What I did is to create a BlockingCollection<T> that will store the events an a Task that will keep taking one event at a time until it will be signaled to stop using a CancellationTokenSource.
My concern is that I'm not handling the synchronization good enough.
This is the class that handles everything (it's used as a WPF ViewModel):
public class EventsTest
{
//private fields
private BlockingCollection<IoEventArgs> _queue;
private CancellationTokenSource _tokenSource;
private IoService _ioService;
private Task _workerTask;
private static EventWaitHandle _eventWaiter;
public EventsTest()
{
_queue = new BlockingCollection<IoEventArgs>();
_tokenSource = new CancellationTokenSource();
_eventWaiter = new EventWaitHandle(false, EventResetMode.AutoReset);
//this is the object that raises multiple events
_ioService = new IoService();
_ioService.IoEvent += _ioService_IoEvent;
//Start Listening
var t = Task.Factory.StartNew(StartListening, _tokenSource, TaskCreationOptions.LongRunning);
}
//IO events listener
private void _ioService_IoEvent(string desc, int portNum)
{
//add events to a blocking collection
_queue.Add(new IoEventArgs() { Description = desc, PortNum = portNum });
}
private void StartListening(object dummy)
{
//process the events one at a time
while (!_tokenSource.IsCancellationRequested)
{
var eve = _queue.Take();
switch (eve.PortNum)
{
case 0:
LongRunningMethod(eve.Description);
break;
case 1:
//invoke a long running method
break;
default:
break;
}
}
}
//sample long running method
private void LongRunningMethod(string data)
{
_eventWaiter.WaitOne(10000);
}
}
How can I make this process more robust in terms of thread safety?
Will adding a lock around each method implementation improve the safety of the process?
Your .Take() won't be canceled so you might wait forever there.
You could pass the token on:
var eve = _queue.Take(_tokenSource);
but then you would have to handle the exception.
A better approach would be the TryTake(out eve, 1000, _tokenSource) and steer with the returned boolean.
Or forget about the CancellationToken and just use AddingComplete()
This sounds like a situation where Microsoft's Reactive Framework is a far better fit.
Your code would look like this:
public class EventsTest
{
private IDisposable _subscription;
public EventsTest()
{
IoService ioService = new IoService();
_subscription =
Observable
.FromEvent<IoEvent, IoEventArgs>(
a => ioService.IoEvent += a, a => ioService.IoEvent -= a)
.Subscribe(eve =>
{
switch (eve.PortNum)
{
case 0:
LongRunningMethod(eve.Description);
break;
case 1:
//invoke a long running method
break;
default:
break;
}
});
}
private void LongRunningMethod(string data)
{
}
}
This should automatically ensure multiple events are queued and will never overlap. If there's problem just drop a .Synchronize() call in before the .Subscribe(...) and then it'll work perfectly.
When you want to cancel the event just call _subscription.Dispose() and it will all be cleaned up for you.
NuGet "Rx-Main" to get the bits you need.
I have looked for solution, but found nothing.
In some class i have event
public class ClassWithEvent
{
public event Action<string> SomeEvent;
...
}
and this event has subscriber
public class SubscriberClass
{
public void SomeMethod(string value)
{
...
}
}
ClassWithEvent objectWithEvent = new ClassWithEvent();
SubscriberClass subscriberObject = new SubscriberClass();
objectWithEvent.SomeEvent += subscriberObject.SomeMethod;
Somewhere in main thread this event can be invoked.
if(SomeEvent != null)
SomeEvent(someString);
And when it is happened its handler has to run in second thread. But every time in the same thread, so this second thread has to be permanent and not be terminated after first execution.
Help me please to implement this.
Create a shared queue:
private BlockingCollection<string> queue = new BlockingCollection<string>();
Create your dedicated thread:
Thread thread = new Thread(HandleEvents);
thread.Start();
Process events on that thread:
private void HandleEvents()
{
foreach (string someValue in queue.GetConsumingEnumerable())
{
//Do stuff
}
}
Place items into the queue when the event is raised:
objectWithEvent.SomeEvent += (x) => queue.Add(x);
On shutdown complete the queue:
queue.CompleteAdding();
Well it's quite tricky but it should work.
You will need to create the permanent thread, for example by starting while(true) there, and a List of actions, which will be checked inside of this loop, something like this:
ConcurrentBag<Action> actions = new ConcurrentBag<Action>();
new Thread(
delegate()
{
while (true)
{
Action a;
if (actions.TryTake(out a))
a();
}
}
).Start();
And every time you want to rise the event just add the action to the List:
actions.Add(new Action(() => { SomeEvent("abc"); }));
Usually this is the stuff i would spend a few hours browsing google and stackoverflow for, however i ran into the problem of how the heck do i word this for a search engine.
I hoping there is a simple way of achieving this, as my current method feels far to "hackish"
What I need to do, if track when several sources of data have completed their loading, and only when all have completed do i load a new view (this is WPF mvvm). Now the data is loaded via a static class termed Repository each one creates a thread and ensure they only a single load operation can happen at once (to avoid multiple threads trying to load into the same collection), each of these classes fires an event called LoadingCompleted when they have finished loading.
I have a single location that loads a large portion of the data (for the first time, there are other locations where the data is reloaded however) what i planned was to hook into each repositories OnLoaded event, and keep track of which have already returned, and when one is returned, mark it as loaded and check to see if any remain. If none remain load the new view, else do nothing.
Something like this:
ShipmentRepository.LoadingComplete += ShipmentRepository_LoadingComplete;
ContainerRepository.LoadingComplete += ContainerRepository_LoadingComplete;
void ContainerRepository_LoadingComplete(object sender, EventArgs e)
{
_containerLoaded = true;
CheckLoaded();
}
void ShipmentRepository_LoadingComplete(object sender, EventArgs e)
{
_shipmentsLoaded = true;
CheckLoaded();
}
private void CheckLoaded()
{
if(_shipmentsLoaded && _containersLoaded && _packagesLoaded)
{
LoadView();
}
}
However as i mentioned this feels clumbsy and hackish, I was hoping there was a cleaner method of doing this.
You can achieve this with Reactive Extensions and using Observable.FromEventPattern in conjunction with the Observable.Zip method. You should be able to do something like:
var shipmentRepositoryLoadingComplete = Observable.FromEventPattern<EventHandler,EventArgs>(h => ShipmentRepository.LoadingComplete += h, h => ShipmentRepository.LoadingComplete -= h);
var containerRepositoryLoadingComplete = Observable.FromEventPattern<EventHandler, EventArgs>(h => ContainerRepository.LoadingComplete += h, h => ContainerRepository.LoadingComplete -= h);
Then you subscibe to the observalbes like this:
var subscription = Observable.Zip(shipmentRepositoryLoadingComplete, containerRepositoryLoadingComplete)
.Subscribe(l => LoadView()));
The subscirption needs to stay alive, so you can save this as a private variable. When both complete events are invoked, the LoadView method should be called. Here is a working console example I used to test this method.
using System;
using System.Reactive.Linq;
namespace RxEventCombine
{
class Program
{
public event EventHandler event1;
public event EventHandler event2;
public event EventHandler event3;
public Program()
{
event1 += Event1Completed;
event2 += Event2Completed;
event3 += Event3Completed;
}
public void Event1Completed(object sender, EventArgs args)
{
Console.WriteLine("Event 1 completed");
}
public void Event2Completed(object sender, EventArgs args)
{
Console.WriteLine("Event 2 completed");
}
public void Event3Completed(object sender, EventArgs args)
{
Console.WriteLine("Event 3 completed");
}
static void Main(string[] args)
{
var program = new Program();
var event1Observable = Observable.FromEventPattern<EventHandler,EventArgs>(h => program.event1 += h, h => program.event1 -= h);
var event2Observable = Observable.FromEventPattern<EventHandler, EventArgs>(h => program.event2 += h, h => program.event2 -= h);
var event3Observable = Observable.FromEventPattern<EventHandler, EventArgs>(h => program.event3 += h, h => program.event3 -= h);
using (var subscription = Observable.Zip(event1Observable, event2Observable, event3Observable)
.Subscribe(l => Console.WriteLine("All events completed")))
{
Console.WriteLine("Invoke event 1");
program.event1.Invoke(null, null);
Console.WriteLine("Invoke event 2");
program.event2.Invoke(null, null);
Console.WriteLine("Invoke event 3");
program.event3.Invoke(null, null);
}
Console.ReadKey();
}
}
}
Output
Invoke event 1
Event 1 completed
Invoke event 2
Event 2 completed
Invoke event 3
Event 3 completed
All events completed
Another way to do this: Add a property LoadingCompleted. For every instance you start a thread return that object to a list. On every loadcompleted set the property to true and in the place you catch the load completed loop through the list (myList.Any(x=>LoadingCompleted == false)) to figure out if everything is completed.
Not the most correct way to do it. But reading your scenario this might do the job.
If you are loading the shipments, containers and packages as asynchronous task then you have several options. As others suggested you can use WhenAll or Join() to wait for all threads to finish before proceeding. However, if your threads have to stay alive and the threads don't stop when they have finished loading, you can use the System.Threading.CountdownEvent as following:
Edit
Added how I would set up the threads and handle the events. Also moved the example from the static Program to an instance, more closely resembeling your situation. Again, if you do not need to do anything in the threads after they have loaded the data, just skip the CountdownEvent altogether and wait for all threads to finish. Much simpler, does not need events and can be achieved using Join() or in this case Task.WaitAll().
class Program
{
static void Main(string[] args)
{
var myWpfObject = new MyWpfObject();
}
}
public class MyWpfObject
{
CountdownEvent countdownEvent;
public MyWpfObject()
{
ShipmentRepository ShipmentRepository = new ShipmentRepository();
ContainerRepository ContainerRepository = new ContainerRepository();
PackageRepository PackageRepository = new PackageRepository();
ShipmentRepository.LoadingComplete += Repository_LoadingComplete;
ContainerRepository.LoadingComplete += Repository_LoadingComplete;
PackageRepository.LoadingComplete += Repository_LoadingComplete;
Task[] loadingTasks = new Task[] {
new Task(ShipmentRepository.Load),
new Task(ContainerRepository.Load),
new Task(PackageRepository.Load)
};
countdownEvent = new CountdownEvent(loadingTasks.Length);
foreach (var task in loadingTasks)
task.Start();
// Wait till everything is loaded.
countdownEvent.Wait();
Console.WriteLine("Everything Loaded");
//Wait till aditional tasks are completed.
Task.WaitAll(loadingTasks);
Console.WriteLine("Everything Completed");
Console.ReadKey();
}
public void Repository_LoadingComplete(object sender, EventArgs e)
{
countdownEvent.Signal();
}
}
And a mock Repository class:
public class ShipmentRepository
{
public ShipmentRepository()
{
}
public void Load()
{
//Simulate work
Thread.Sleep(1000);
if (LoadingComplete != null)
LoadingComplete(this, new EventArgs());
Console.WriteLine("Finished loading shipments");
DoAditionalWork();
}
private void DoAditionalWork()
{
//Do aditional work after everything is loaded
Thread.Sleep(5000);
Console.WriteLine("Finished aditional shipment work");
}
public event EventHandler LoadingComplete;
}
This is a fictional example but I was wandering what happens if the InitialiseTimer function gets called twice. Does the timer elapsed function get triggered twice. Will this change if the functions are made static?
private static void InitialiseTimer()
{
TheTimer = new System.Timers.Timer();
TheTimer.Interval = 400;
TheTimer.Elapsed += new ElapsedEventHandler(TheTimer_Elapsed);
TheTimer.AutoReset = false;
}
public void TheTimer_Elapsed(object sender, ElapsedEventArgs e)
{
//Do stuff in here
}
I was going to use below to prevent this
Has an event handler already been added?
Thanks,
Richard
If you register the event handler twice, it will be invoked twice every time the event is raised.
This won't change if you make TheTimer_Elapsed static, because you'll still hold two references to this static method.
In most cases there's no need to write compicated things like what Blair Conrad posted in the question you linked to. Just don't forget to use -= every time you have += and you'll be safe.
I think the following demonstrates the scenario and does indeed fire twice, also propose a simple change (commented code) to the Init method that should fix the behavior. (Not thread safe btw, additional locks would be required)
[TestClass]
public class UnitTest1
{
[TestMethod]
public void TestMethod1()
{
var counter = 0;
var ts = new ThreadStart(() =>
{
Foo.Fired += (o, e) =>
{
counter++;
};
Foo.InitialiseTimer();
Foo.InitialiseTimer();
});
var t = new Thread(ts);
t.Start();
Thread.Sleep(30);
Assert.AreEqual(1, counter);
}
}
public class Foo
{
private static System.Timers.Timer TheTimer = null;
public static event EventHandler Fired;
public static void InitialiseTimer()
{
//if (TheTimer != null)
//{
// TheTimer.Stop();
// TheTimer = null;
//}
TheTimer = new System.Timers.Timer();
TheTimer.Interval = 10;
TheTimer.Elapsed += new ElapsedEventHandler(TheTimer_Elapsed);
TheTimer.AutoReset = false;
TheTimer.Start();
}
public static void TheTimer_Elapsed(object sender, ElapsedEventArgs e)
{
//Do stuff in here
if (Fired != null)
{
Fired(null, null);
}
}
}
if you call the method InitialiseTimer twice you will create two Timers each of them will have only one event handler attached but they might elapse both. It's not really about having the method static or not, it's more about the method itself, you could check if TheTimer is null and do the rest only if it's null so you assign it only once.
If event is registered twice you will have two executions.
You can check if event is null, and the problem will be solved.
Static or not, you are recreating the Timer. So you can invoke the InitialiseTimer many, many times without adding more than a single handler. You will end up with many timers though...