我正在尝试学习新的。net响应式扩展框架，它听起来像是我应用程序的完美解决方案。在学习了示例之后(LINQ仍然相当薄弱)，我正在努力找出如何利用RX框架来完成下面的任务。
目标是在自定义数据源和GUI之间创建一个可配置的事件"中继"。中继将使用LINQ来测试和过滤传入的事件，在等待下一个时间间隔时将合格的事件缓存在列表中，然后同步到GUI线程并按照接收到的顺序播放事件。
如何使用RX来协调缓存、过滤和中继多个事件，例如用于从自定义数据源添加、更改和删除值的事件?
这可能要求很多，但是任何关于如何处理这个问题的指导都将非常感激。
public delegate void EventDelegateAdd(Thing thing);
public delegate void EventDelegateChange(Thing thing);
public delegate void EventDelegateRemove(Thing thing);
public delegate void EventDelegateBulkChangesStart();
public delegate void EventDelegateBulkChangesEnd();
// The "Things" that are stored in MyCustomDataSource
public class Thing
{
public string Key { get; set; }
public string Title { get; set; }
public object OtherStuff { get; set; }
}
// A custom observable data source with events that indicate when Things are
// added, changed, or removed.
public class MyCustomDataSource
{
public event EventDelegateAdd AddingThing;
public event EventDelegateChange ChangingThing;
public event EventDelegateRemove RemovingThing;
// The rest of the class that manages the database of Things...
}
// This class forms a configurable event bridge between the MyCustomDataSource and
// the GUI. The goal is to cache, filter, and throttle the events so that the GUI
// updates only occasionally with bulk changes that are relevant for that control.
public class MyEventCachingBridge
{
private MyCustomDataSource mSource;
public event EventDelegateAdd AddingThing;
public event EventDelegateChange ChangingThing;
public event EventDelegateRemove RemovingThing;
public event EventDelegateBulkChangesStart BulkChangesStart;
public event EventDelegateBulkChangesEnd BulkChangesEnd;
public MyEventCachingBridge(MyCustomDataSource source, int eventRelayInterval)
{
mSource = source;
// Magical Reactive Extensions code goes here that subscribes to all 3 events...
//
// mSource.AddingThing
// mSource.ChangingThing
// mSource.RemovingThing
//
// ...filters and records a list of the events as they are received ( maintaining order of events too ),
// then every eventRelayInterval milliseconds, plays back the events in bulk to update the GUI
// ( on the GUIs thread ). Note that LINQ will be used to filter the Things so that a subset of
// Thing changes are relayed to the GUI - i.e. - not all Thing events are observed by the GUI.
}
public void PlayBackCachedEvents()
{
BulkChangesStart(); // Raise Event to notify GUI to suspend screen updates
// Play back the list of events to push changes to ListView, TreeView, graphs, etc...
//
// this.AddingThing(Thing); // Fire events in order received
// this.ChangingThing(Thing); // Fire events in order received
// this.RemovingThing(Thing); // Fire events in order received
BulkChangesEnd(); // Raise Event to notify GUI to allow control refresh
}
给定所描述的任务，我不确定在一般类代码中应该协调什么，以及应该在RX语句中隐藏什么。
我也很欣赏这样一个事实:将3个事件合并成一个事件，并使用一个枚举来指示事件的目的，这将极大地简化事情，但是在实际的应用程序中需要缓存的事件远不止3个。我不希望使用恒定的Switch语句测试来识别每个事件。将会有大量的事件被路由到可能众多的GUI界面。
谢谢你的建议
啊，我喜欢玩Rx…这是一种方法;首先是部分，然后是全部:
编辑:修改以适应注释
首先，你需要为事件设置流——当我们在做这件事的时候，让我们用Rx的"新特性"替换桥暴露的"旧".NET事件模式:
public Subject<Thing> BufferedAdds {get; private set;}
public Subject<Thing> BufferedChanges {get; private set;}
public Subject<Thing> BufferedRemoves {get; private set;}
_adds = Observable.FromEvent<EventDelegateAdd, Thing>(
ev => new EventDelegateAdd(ev),
h => mSource.AddingThing += h,
h => mSource.AddingThing -= h);
BufferedAdds = new Subject<Thing>();
_changes = Observable.FromEvent<EventDelegateChange, Thing>(
ev => new EventDelegateChange(ev),
h => mSource.ChangingThing += h,
h => mSource.ChangingThing -= h);
BufferedChanges = new Subject<Thing>();
_removes = Observable.FromEvent<EventDelegateRemove, Thing>(
ev => new EventDelegateRemove(ev),
h => mSource.RemovingThing += h,
h => mSource.RemovingThing -= h);
BufferedRemoves = new Subject<Thing>();
我们现在也要把所有的东西都和IScheduler联系起来——IScheduler的使用是测试这个该死的东西的一个关键方面，而不是到处都是Thread.Sleep——我强烈建议研究这个话题!
public MyEventCachingBridge(
MyCustomDataSource source,
int eventRelayInterval,
IScheduler scheduler)
{
然后，您将希望将所有传入的事件合并到单个流中，然后根据时间窗口对该流进行"块"—Buffer操作符是理想的:
_buffer = Observable.Merge(scheduler,
_adds.Select(e => Tuple.Create(e, ThingEventType.Add)),
_changes.Select(e => Tuple.Create(e, ThingEventType.Change)),
_removes.Select(e => Tuple.Create(e, ThingEventType.Remove)))
.Buffer(TimeSpan.FromMilliseconds(eventRelayInterval), scheduler);
注意，我正在将事件的类型打包回流中——这样我们就可以在播放期间采取适当的操作——枚举是:
private enum ThingEventType
{
Add,
Change,
Remove
}
List与同步:
private Queue<IList<Tuple<Thing,ThingEventType>>> _eventQueue;
private static object SyncRoot = new object();
_eventQueue = new Queue<IList<Tuple<Thing,ThingEventType>>>();
// A serial disposable is a sort of "Disposable holder" - when you change it's
// Disposable member, it auto-disposes what you originally had there...no real
// need for it here, but potentially useful later
_watcherDisposable = new SerialDisposable();
_watcherDisposable.Disposable = _buffer
.ObserveOn(_scheduler)
.Subscribe(batch =>
{
lock(SyncRoot) { _eventQueue.Enqueue(batch); }
});
_disposables.Add(_watcherDisposable);
让我们也"自动连接"播放突发脉冲每eventRelayInterval毫秒:
var pulse = Observable.Interval(
TimeSpan.FromMilliseconds(eventRelayInterval),
_scheduler);
_disposables.Add(pulse
.ObserveOn(_scheduler)
.Subscribe(x => PlayBackCachedEvents()));
订阅总是可丢弃的，你会想要丢弃它们，所以让我们为此添加一些东西:
public class MyEventCachingBridge : IDisposable
{
CompositeDisposable _disposables;
public void Dispose()
{
_disposables.Dispose();
}
现在播放:
public void PlayBackCachedEvents()
{
BulkChangesStart(); // Raise Event to notify GUI to suspend screen updates
// Play back the list of events to push changes
lock(SyncRoot)
{
foreach(var batch in _eventQueue)
{
// Play back the list of events to push changes to ListView, TreeView, graphs, etc...
foreach(var evt in batch)
{
switch(evt.Item2)
{
case ThingEventType.Add: BufferedAdds.OnNext(evt.Item1); break;
case ThingEventType.Change: BufferedChanges.OnNext(evt.Item1);break;
case ThingEventType.Remove: BufferedRemoves.OnNext(evt.Item1);break;
}
}
}
_eventQueue.Clear();
}
BulkChangesEnd(); // Raise Event to notify GUI to allow control refresh
}
现在-我们想要在消费者端也很花哨，所以让我们模拟一个UI窗口(这是WPF，相应调整):
public class BridgeConsumer : Window, IDisposable
{
private readonly CompositeDisposable _disposables;
private IScheduler _scheduler;
private StackPanel _panel;
public void OnLoaded(object sender, RoutedEventArgs ea)
{
_panel = new StackPanel();
this.Content = _panel;
}
public BridgeConsumer(MyEventCachingBridge bridge, IScheduler scheduler)
{
// for cleanup of any subscriptions
_disposables = new CompositeDisposable();
_disposables.Add(bridge);
_scheduler = scheduler;
Loaded += OnLoaded;
// setup a listener for the bulk start/end events on the bridge
var bulkStart = Observable.FromEvent(
h => bridge.BulkChangesStart += new EventDelegateBulkChangesStart(h),
h => bridge.BulkChangesStart -= new EventDelegateBulkChangesStart(h));
var bulkEnd = Observable.FromEvent(
h => bridge.BulkChangesEnd += new EventDelegateBulkChangesEnd(h),
h => bridge.BulkChangesEnd -= new EventDelegateBulkChangesEnd(h));
// the "meaty bit" -
// 1. create a "window" defined by bulk start/end events
// 2. inside that "window", trap any occurrences on a
// merged view of adds/changes/removes
// 3. foreach event in that window, select that event
// (i.e., give us window contents as a stream of sorts)
var bridgeWatcher =
from thingEventWindow in
Observable.Merge(
bridge.BufferedAdds.Select(t => Tuple.Create("add", t)),
bridge.BufferedChanges.Select(t => Tuple.Create("change", t)),
bridge.BufferedRemoves.Select(t => Tuple.Create("remove", t))
)
.Window(bulkStart, start => bulkEnd)
from thingEvent in thingEventWindow
select thingEvent;
// this could just as easily be a method, a bound call to the viewmodel, etc
Action<Thing, string, DateTimeOffset> addToList = (thing, msg, ts) =>
{
var text = new TextBlock()
{
Text = string.Format(
"At:{0} Key:{2} Msg:{3} - nowTime = {1}",
thing.TimeStamp,
ts,
thing.Key,
msg) };
_panel.Children.Add(text);
};
_disposables.Add(bridgeWatcher
// CAREFUL! "ObserveOn" means what you'd think "SubscribeOn" would
.ObserveOnDispatcher()
.Subscribe(tup =>
{
addToList(tup.Item2, tup.Item1, _scheduler.Now);
}));
}
public void Dispose()
{
// clean up
if(_disposables != null) _disposables.Dispose();
}
}
整件事:
void Main()
{
var scheduler = Scheduler.Default;
var rnd = new Random();
var canceller = new CancellationTokenSource();
var source = new MyCustomDataSource();
var eventRelayInterval = 2000;
var bridge = new MyEventCachingBridge(source, eventRelayInterval, scheduler);
var window = new BridgeConsumer(bridge);
window.Closed += (o,e) => { canceller.Cancel(); window.Dispose(); };
window.Show();
Task.Factory.StartNew(
() =>
{
while(true)
{
var thing = new Thing()
{
Key = "added thing " + rnd.Next(0, 100),
Title = "title for added thing",
TimeStamp = scheduler.Now.DateTime
};
source.FireAdd(thing);
Thread.Sleep(rnd.Next(1,10) * 100);
}
}, canceller.Token);
}
public class BridgeConsumer : Window, IDisposable
{
private readonly CompositeDisposable _disposables;
private StackPanel _panel;
public void OnLoaded(object sender, RoutedEventArgs ea)
{
_panel = new StackPanel();
this.Content = _panel;
}
public BridgeConsumer(MyEventCachingBridge bridge)
{
_disposables = new CompositeDisposable();
_disposables.Add(bridge);
Loaded += OnLoaded;
var bulkStart = Observable.FromEvent(
h => bridge.BulkChangesStart += new EventDelegateBulkChangesStart(h),
h => bridge.BulkChangesStart -= new EventDelegateBulkChangesStart(h));
var bulkEnd = Observable.FromEvent(
h => bridge.BulkChangesEnd += new EventDelegateBulkChangesEnd(h),
h => bridge.BulkChangesEnd -= new EventDelegateBulkChangesEnd(h));
var bridgeWatcher =
from thingEventWindow in
Observable.Merge(
bridge.BufferedAdds.Select(t => Tuple.Create("add", t)),
bridge.BufferedChanges.Select(t => Tuple.Create("change", t)),
bridge.BufferedRemoves.Select(t => Tuple.Create("remove", t))
)
.Window(bulkStart, start => bulkEnd)
from thingEvent in thingEventWindow
select thingEvent;
Action<Thing, string> addToList = (thing, msg) =>
{
var text = new TextBlock()
{
Text = string.Format(
"At:{0} Key:{1} Msg:{2}",
thing.TimeStamp,
thing.Key,
msg)
};
_panel.Children.Add(text);
};
_disposables.Add(bridgeWatcher.ObserveOnDispatcher().Subscribe(tup =>
{
addToList(tup.Item2, tup.Item1);
}));
}
public void Dispose()
{
if(_disposables != null) _disposables.Dispose();
}
}
public delegate void EventDelegateAdd(Thing thing);
public delegate void EventDelegateChange(Thing thing);
public delegate void EventDelegateRemove(Thing thing);
public delegate void EventDelegateBulkChangesStart();
public delegate void EventDelegateBulkChangesEnd();
// The "Things" that are stored in MyCustomDataSource
public class Thing
{
public DateTime TimeStamp {get; set;}
public string Key { get; set; }
public string Title { get; set; }
public object OtherStuff { get; set; }
public override string ToString()
{
return string.Format("At:{0} Key:{1} Title:{2}", this.TimeStamp, this.Key, this.Title);
}
}
// A custom observable data source with events that indicate when Things are
// added, changed, or removed.
public class MyCustomDataSource
{
public event EventDelegateAdd AddingThing = delegate { };
public event EventDelegateChange ChangingThing = delegate { };
public event EventDelegateRemove RemovingThing = delegate { };
// The rest of the class that manages the database of Things...
public void FireAdd(Thing toAdd)
{
AddingThing(toAdd);
}
public void FireChange(Thing toChange)
{
ChangingThing(toChange);
}
public void FireRemove(Thing toRemove)
{
RemovingThing(toRemove);
}
}
// This class forms a configurable event bridge between the MyCustomDataSource and
// the GUI. The goal is to cache, filter, and throttle the events so that the GUI
// updates only occasionally with bulk changes that are relevant for that control.
public class MyEventCachingBridge : IDisposable
{
private enum ThingEventType
{
Add,
Change,
Remove
}
private MyCustomDataSource mSource;
private IScheduler _scheduler;
public event EventDelegateBulkChangesStart BulkChangesStart = delegate { };
public event EventDelegateBulkChangesEnd BulkChangesEnd = delegate { };
public IObservable<Thing> RawAdds {get; private set;}
public IObservable<Thing> RawChanges {get; private set;}
public IObservable<Thing> RawRemoves {get; private set;}
public Subject<Thing> BufferedAdds {get; private set;}
public Subject<Thing> BufferedChanges {get; private set;}
public Subject<Thing> BufferedRemoves {get; private set;}
private readonly IObservable<IList<Tuple<Thing,ThingEventType>>> _buffer;
private List<IList<Tuple<Thing,ThingEventType>>> _eventQueue;
private static object SyncRoot = new object();
private readonly CompositeDisposable _disposables;
private readonly SerialDisposable _watcherDisposable;
public MyEventCachingBridge(MyCustomDataSource source, int eventRelayInterval, IScheduler scheduler)
{
_disposables = new CompositeDisposable();
mSource = source;
_scheduler = scheduler;
_eventQueue = new List<IList<Tuple<Thing,ThingEventType>>>();
// Magical Reactive Extensions code goes here that subscribes to all 3 events...
//
// mSource.AddingThing
// mSource.ChangingThing
// mSource.RemovingThing
//
// ...filters and records a list of the events as they are received ( maintaining order of events too ),
// then every eventRelayInterval milliseconds, plays back the events in bulk to update the GUI
// ( on the GUIs thread ). Note that LINQ will be used to filter the Things so that a subset of
// Thing changes are relayed to the GUI - i.e. - not all Thing events are observed by the GUI.
RawAdds = Observable.FromEvent<EventDelegateAdd, Thing>(
ev => new EventDelegateAdd(ev),
h => mSource.AddingThing += h,
h => mSource.AddingThing -= h);
BufferedAdds = new Subject<Thing>();
RawChanges = Observable.FromEvent<EventDelegateChange, Thing>(
ev => new EventDelegateChange(ev),
h => mSource.ChangingThing += h,
h => mSource.ChangingThing -= h);
BufferedChanges = new Subject<Thing>();
RawRemoves = Observable.FromEvent<EventDelegateRemove, Thing>(
ev => new EventDelegateRemove(ev),
h => mSource.RemovingThing += h,
h => mSource.RemovingThing -= h);
BufferedRemoves = new Subject<Thing>();
_buffer = Observable.Merge(
_scheduler,
RawAdds.Select(e => Tuple.Create(e, ThingEventType.Add)),
RawChanges.Select(e => Tuple.Create(e, ThingEventType.Change)),
RawRemoves.Select(e => Tuple.Create(e, ThingEventType.Remove)))
.Buffer(TimeSpan.FromMilliseconds(eventRelayInterval), _scheduler);
_watcherDisposable = new SerialDisposable();
_watcherDisposable.Disposable = _buffer
.ObserveOn(_scheduler)
.Subscribe(batch =>
{
lock(SyncRoot) { _eventQueue.Add(batch); }
});
_disposables.Add(_watcherDisposable);
var pulse = Observable.Interval(TimeSpan.FromMilliseconds(eventRelayInterval), _scheduler);
_disposables.Add(pulse.ObserveOn(_scheduler).Subscribe(x => PlayBackCachedEvents()));
}
private void PlayBackCachedEvents()
{
BulkChangesStart(); // Raise Event to notify GUI to suspend screen updates
try
{
//_eventQueue.Dump();
lock(SyncRoot)
{
foreach(var batch in _eventQueue)
{
// Play back the list of events to push changes to ListView, TreeView, graphs, etc...
foreach(var evt in batch)
{
switch(evt.Item2)
{
case ThingEventType.Add: BufferedAdds.OnNext(evt.Item1); break;
case ThingEventType.Change: BufferedChanges.OnNext(evt.Item1);break;
case ThingEventType.Remove: BufferedRemoves.OnNext(evt.Item1);break;
}
}
}
_eventQueue.Clear();
}
}
catch(Exception ex)
{
Console.WriteLine("Exception during playback:" + ex);
}
BulkChangesEnd(); // Raise Event to notify GUI to allow control refresh
}
public void Dispose()
{
_disposables.Dispose();
}
}
我对这个特别的展示有点晚了，但我认为你需要的是一个可观察的缓存/列表，它将提供所有添加，更新和删除的更改集。
我已经实现了这一点，并使其开源:
博客:http://dynamicdataproject.wordpress.com/
演示:https://github.com/RolandPheasant/TradingDemo