WCF Duplex client idle wont update userinterface - wpf

I have clients thats is connected to a duplex wcf service 24/6 it is restarted every sunday.
In the client iam using a listview to display some information.
The listview itemssource is binded to a custom ObservableCollection.
The client is calling a keepalive method every minute to the wcf service.
My problem here is the client works fine when there is activity in the client. But when there is no activity and the application just run the keepalive method for about 10-16 hours. And iam trying to add and remove data to the listview it seems that nothing works. But the wcf service logging the method add and remove is working fine. Its like the userinterface isnt updating. When i restart the application everything works fine.
how do i solve this problem ?
My custom ObservableCollection object code :
public class ObservableOrderResponseQueue : INotifyCollectionChanged, IEnumerable<OrderResponse>
{
public event NotifyCollectionChangedEventHandler CollectionChanged = (o, e) => { };
private List<OrderResponse> _list = new List<OrderResponse>();
/// <summary>
/// Adds to the list.
/// </summary>
/// <param name="orderResponse">OrderResponse.</param>
public void Add(OrderResponse orderResponse)
{
//Only 6 items in list is possible if more then 6 remove the first item.
if (_list.Count >= 6)
{
RemoveAt(0);
}
this._list.Add(orderResponse);
CollectionChanged(this, new NotifyCollectionChangedEventArgs(NotifyCollectionChangedAction.Add, orderResponse, (_list.Count - 1)));
}
/// <summary>
/// Remove from list.
/// </summary>
/// <param name="index">Item index to remove.</param>
public void RemoveAt(int index)
{
OrderResponse order = this._list[index];
this._list.RemoveAt(index);
CollectionChanged(this, new NotifyCollectionChangedEventArgs(NotifyCollectionChangedAction.Remove, order, index));
}
/// <summary>
/// Remove from list.
/// </summary>
/// <param name="orderResponse">Item to be removed.</param>
public void Remove(OrderResponse orderResponse)
{
if (_list.Count == 0) return;
var item = _list.Where(o => o.OrderDetail.TrayCode == orderResponse.OrderDetail.TrayCode).FirstOrDefault();
int index = _list.IndexOf(item);
if (index == -1) return;
CollectionChanged(this, new NotifyCollectionChangedEventArgs(NotifyCollectionChangedAction.Remove, item, index));
this._list.RemoveAt(index);
}
#region IEnumerable<OrderResponse> Members
public IEnumerator<OrderResponse> GetEnumerator()
{
return _list.GetEnumerator();
}
#endregion
#region IEnumerable Members
IEnumerator IEnumerable.GetEnumerator()
{
return _list.GetEnumerator();
}
#endregion
}
Here is how i bind the usercontrol to the class.
//Set up client callbacks events
App.clientBack.ClientNotified += new ClientNotifiedEventHandler(clientBack_ClientNotified);
App.clientBack.AddToDisplayEvent += new AddToDisplayEventHandler(clientBack_AddToDisplayEvent);
App.clientBack.RemoveFromDisplayEvent += new RemoveFromDisplayEventHandler(clientBack_RemoveFromDisplayEvent);
App.clientBack.UpdateQueueDisplayEvent += new UpdateQueueDisplayEventHandler(clientBack_UpdateQueueDisplayEvent);
//Show one chair or many.
if (_settings.IsOneChair)
{
userControlOneChair.ItemSource = _queueProductionItems;
}
else
{
userControlChairs.ItemsSource = _queueProductionItems;
}
Remove and add methods
void clientBack_RemoveFromDisplayEvent(object sender, RemoveFromDisplayEventArgs e)
{
try
{
_logger.Info("Remove from display.");
userControlChairs.Dispatcher.Invoke((Action)(() =>
{
_queueProductionItems.Remove(e.OrderResponse);
}));
}
catch (Exception ex)
{
MessageBox.Show(ex.Message);
}
}
void clientBack_AddToDisplayEvent(object sender, AddToDisplayEventArgs e)
{
try
{
_logger.Info("Add to display.");
userControlChairs.Dispatcher.Invoke((Action)(() =>
{
_queueProductionItems.Add(e.OrderResponse);
}));
}
catch (Exception ex)
{
MessageBox.Show(ex.Message);
}
}
Thanks for help!

What i did was implementing a Heartbeat mechanism. And it all worked out.

Related

How to track number of spam clicks on a button in WPF

I have a button that skips a video on by x seconds. if a user spam clicks that button my video updates over and over again which is an expensive operation. What is the best way to stop a user spamming the button? I am using a routed ui command and want to add up the seconds and do 1 operation. Is a delay timer best practice here? delay the operation for 10ms and reset the delay on every click? or is there something built into wpf that can help?
UPDATE:
I would like to track the number of clicks a user is making during the spam click of the button
I really hope the async way works, while we were trying that out i created a solution, feel free to tell me all i did wrong and any bad practices.
I decided to use dispatcher timer for this even though i didn't really want to. couldn't find any better practices online.
private TimeSpan overallSkipSpeed = TimeSpan.Zero;
private readonly TimeSpan Interval = TimeSpan.FromMilliseconds(400);
private DispatcherTimer _dispatcherTimer;
private TimeSpan _time;
// Execute command function
public void ExecuteClickCommand()
{
// If the timer isn't going create and start it
if (_dispatcherTimer == null)
{
overallSkipSpeed = TimeSpanToModifyBy(skipSpeed, skipForward);
_time = Interval;
_dispatcherTimer = new DispatcherTimer(Interval, DispatcherPriority.Normal, Tick, Application.Current.Dispatcher);
_dispatcherTimer.Start();
}
else // If the timer is going reset to interval
{
// THIS IS WHERE I ADDED MY SKIP VALUES TOGETHER
// So value from last click + value from this click
_dispatcherTimer.Stop();
_time = Interval;
_dispatcherTimer.Start();
}
}
// Method to run when timer ticks over
private void Tick(object sender, EventArgs eventArgs)
{
// if the timer has reached below zero
if (_time <= TimeSpan.Zero)
{
_dispatcherTimer.Stop();
_dispatcherTimer = null;
_time = TimeSpan.FromSeconds(0);
// HERE IS WHERE WE CAN NOW SKIP VIDEO BY
// THE SKIP SPEED WE HAVE ACCUMULATED
}
else
{
_time = _time.Add(-Interval);
}
}
I have gone one step further with this and created my own command.
This command works like a relay command but will delay if you set a delay time span in initialisation. you can retrieve the number of times it was clicked in your execute method.
Initialise the command:
ICommand DelayedClickCommand = new DelayedCommand(ExecuteDelayedClickCommand, TimeSpan.FromMilliseconds(200));
Create an execute method and retrive the amount of times clicked:
private void ExecuteClickCommand()
{
TimesClicked = ((DelayedCommand)ClickCommand).TimesClicked;
}
and here is the command class:
public class DelayedCommand : ICommand
{
private readonly Action _methodToExecute;
private readonly Func<bool> _canExecuteEvaluator;
private readonly DispatcherTimer _dispatcherTimer;
public int TimesClicked;
public event EventHandler CanExecuteChanged
{
add { CommandManager.RequerySuggested += value; }
remove { CommandManager.RequerySuggested -= value; }
}
/// <summary>
/// A command to stop the spamming of the <see cref="Execute"/> method
/// </summary>
/// <param name="methodToExecute">Method to run when command executes</param>
/// <param name="canExecuteEvaluator">Method used to determine if the command can execute</param>
/// <param name="delayTime">The cool down period required between click execution</param>
public DelayedCommand(Action methodToExecute, Func<bool> canExecuteEvaluator, TimeSpan delayTime)
{
_methodToExecute = methodToExecute;
_canExecuteEvaluator = canExecuteEvaluator;
_dispatcherTimer = new DispatcherTimer(delayTime, DispatcherPriority.Normal, Callback, Application.Current.Dispatcher);
}
/// <summary>
/// A command to stop the spamming of the <see cref="Execute"/> method
/// when no <see cref="CanExecute"/> method is required
/// </summary>
/// <param name="methodToExecute">Method to run when command executes</param>
/// <param name="delayTime">The cool down period required between click execution</param>
public DelayedCommand(Action methodToExecute, TimeSpan delayTime)
: this(methodToExecute, null, delayTime)
{
}
/// <summary>
/// A command when only a <see cref="Execute"/> method is needed
/// </summary>
/// <param name="methodToExecute">Method to run when command executes</param>
public DelayedCommand(Action methodToExecute)
: this(methodToExecute, null, TimeSpan.Zero)
{
}
/// <summary>
/// A command taking a <see cref="Execute"/> Method and a <see cref="CanExecute"/> method
/// </summary>
/// <param name="methodToExecute">Method to run when command executes</param>
/// <param name="canExecuteEvaluator">Method used to determine if the command can execute</param>
public DelayedCommand(Action methodToExecute, Func<bool> canExecuteEvaluator)
: this(methodToExecute, canExecuteEvaluator, TimeSpan.Zero)
{
}
public bool CanExecute(object parameter)
{
if (_canExecuteEvaluator == null)
{
return true;
}
return _canExecuteEvaluator.Invoke();
}
public void Execute(object parameter)
{
if (!_dispatcherTimer.IsEnabled)
TimesClicked = 0;
TimesClicked++;
_dispatcherTimer?.Stop();
_dispatcherTimer?.Start();
}
private void Callback(object sender, EventArgs eventArgs)
{
_dispatcherTimer.Stop();
_methodToExecute.Invoke();
}
}
Note: that when you spam click this command execute will not run untill 200ms after the last click was performed, giving a lagging effect. I have added a sample project to git hub and will add better commands for this question on there
https://github.com/sgreaves1/DelayedCommands
Guess i'm the laziest person here...
public class PostponeCommand : ICommand
{
private TimeSpan _delay;
private Action<object> _command;
private CancellationTokenSource _cancellation;
public PostponeCommand(Action<object> command, int delayMs)
{
this._command = command;
this._delay = TimeSpan.FromMilliseconds(delayMs);
}
public bool CanExecute(object parameter)
{
return true;
}
public async void Execute(object parameter)
{
_cancellation?.Cancel();
_cancellation = new CancellationTokenSource();
try
{
await Task.Delay(_delay, _cancellation.Token);
_command?.Invoke(parameter);
}
catch (TaskCanceledException ex)
{
// canceled
}
}
public event EventHandler CanExecuteChanged;
}
Not sure about build-in Command ability to do this, but you can do it with delay (updated based on comments):
private int spamCount = 0;
private int delayValue = 0;
private object isHoldedLock = new object();
private bool isHolded = false;
public bool CanProceed(int delay, Action updateVideo)
{
lock (this.isHoldedLock)
{
if (this.isHolded)
{
this.spamCount++;
this.delayValue = delay;
return false;
}
this.isHolded = true;
this.delayValue = delay;
Task.Run(async () =>
{
while (this.delayValue > 0)
{
await Task.Delay(100);
this.delayValue -= 100;
}
updateVideo();
lock (this.isHoldedLock)
{
this.isHolded = false;
}
});
return true;
}
}
Process/reset spamCount value inside SkipVideo any way you need.
And using in your command handler:
private void InvokedFromCommand()
{
if (CanProceed(1000, SkipVideo()))
{
// SkipVideo();
}
}

MVVM bind RelayCommand CanExecute to a Property?

I have a Timer and three buttons to control it: Start, Stop, and Pause.
Each button is bound to a RelayCommand.
I have a TimerState property of type enum TimerState. (This is useful for setting various GUI elements.)
Is there a way to somehow bind the RelayCommands' CanExecute functionality to the TimerState property?
Currently, I have 3 methods that look like this:
private bool CanStartTimer()
{
return (TimerState == TimerState.Stopped || TimerState == TimerState.Paused);
}
In the TimerState setter, I call
StartTimerCmd.RaiseCanExecuteChanged();
Is there a better way bind the CanExecute state of the RelayCommands to a property like TimerState?
Thanks for any insight.
I've implemented a class to handle commands, actually it's based on DelegateCommand because i'm using PRISM but it could easily be changed to be used with RelayCommand or any other class implementing ICommand
It could have bugs, i've not yet fully tested it, however it works fine in my scenarios, here it is:
public class MyDelegateCommand<TViewModel> : DelegateCommand where TViewModel : INotifyPropertyChanged {
private List<string> _PropertiesToWatch;
public MyDelegateCommand(TViewModel viewModelInstance, Action executedMethod)
: base(executedMethod) {
}
public MyDelegateCommand(TViewModel viewModelInstance, Action executedMethod, Func<bool> canExecuteMethod)
: base(executedMethod, canExecuteMethod) {
}
/// <summary>
///
/// </summary>
/// <param name="viewModelInstance"></param>
/// <param name="executedMethod"></param>
/// <param name="selector"></param>
public MyDelegateCommand(TViewModel viewModelInstance, Action executedMethod, Func<bool> canExecuteMethod, Expression<Func<TViewModel, object>> propertiesToWatch)
: base(executedMethod, canExecuteMethod) {
_PropertiesToWatch = RegisterPropertiesWatcher(propertiesToWatch);
viewModelInstance.PropertyChanged += PropertyChangedHandler;
}
/// <summary>
/// handler that, everytime a monitored property changes, calls the RaiseCanExecuteChanged of the command
/// </summary>
/// <param name="sender"></param>
/// <param name="e"></param>
private void PropertyChangedHandler(object sender, PropertyChangedEventArgs e) {
if (_PropertiesToWatch.Contains(e.PropertyName)) {
this.OnCanExecuteChanged();
}
}
/// <summary>
/// giving an expression that identify a propriety or a list of properties, return the property names obtained from the expression
/// Examples on selector usage
/// proprietà singola:
/// entity => entity.PropertyName
/// proprietà multiple
/// entity => new { entity.PropertyName1, entity.PropertyName2 }
/// </summary>
/// <param name="selector"></param>
/// <returns></returns>
protected List<string> RegisterPropertiesWatcher(Expression<Func<TViewModel, object>> selector) {
List<string> properties = new List<string>();
System.Linq.Expressions.LambdaExpression lambda = (System.Linq.Expressions.LambdaExpression)selector;
if (lambda.Body is System.Linq.Expressions.MemberExpression) {
System.Linq.Expressions.MemberExpression memberExpression = (System.Linq.Expressions.MemberExpression)(lambda.Body);
properties.Add(memberExpression.Member.Name);
}
else if (lambda.Body is System.Linq.Expressions.UnaryExpression) {
System.Linq.Expressions.UnaryExpression unaryExpression = (System.Linq.Expressions.UnaryExpression)(lambda.Body);
properties.Add(((System.Linq.Expressions.MemberExpression)(unaryExpression.Operand)).Member.Name);
}
else if (lambda.Body.NodeType == ExpressionType.New) {
NewExpression newExp = (NewExpression)lambda.Body;
foreach (var argument in newExp.Arguments) {
if (argument is System.Linq.Expressions.MemberExpression) {
System.Linq.Expressions.MemberExpression mExp = (System.Linq.Expressions.MemberExpression)argument;
properties.Add(mExp.Member.Name);
}
else {
throw new SyntaxErrorException("Syntax Error, selector has to be an expression that returns a new object containing a list of properties, e.g.: s => new { s.Property1, s.Property2 }");
}
}
}
else {
throw new SyntaxErrorException("Syntax Error, selector has to be an expression that returns a new object containing a list of properties, e.g.: s => new { s.Property1, s.Property2 }");
}
return properties;
}
}
note that my solution implies that this command has to be wired with the viewmodel that handle it and the viewmodel has to implement the INotifyPropertyChanged interface.
first two constructor are there so the command is backward compatible with DelegateCommand but the 3rd is the important one, that accepts a linq expression to specify which property to monitor
the usage is pretty simple and easy to understand, let me write it here with methods but of course you can create your handler methods. Suppose you have have a ViewModel called MyViewModel with two properties (PropertyX and PropertyY) that rise the propertychanged event, and somewhere in it you create an instance of SaveCommand, it would look like this:
this.SaveCommand = new MyDelegateCommand<MyViewModel>(this,
//execute
() => {
Console.Write("EXECUTED");
},
//can execute
() => {
Console.Write("Checking Validity");
return PropertyX!=null && PropertyY!=null && PropertyY.Length < 5;
},
//properties to watch
(p) => new { p.PropertyX, p.PropertyY }
);
maybe i'll create an article somewhere about this, but this snippet should be clear i hope
Fabio's answer works well. Here's a parameterized version for DelegateCommand<T>. (I've tightened up the code a little, too.)
public class DepedencyCommand<TViewModel, TArg> : DelegateCommand<TArg>
where TViewModel : INotifyPropertyChanged
{
private readonly List<string> _propertiesToWatch;
public DepedencyCommand(Action<TArg> executedMethod)
: base(executedMethod) { }
public DepedencyCommand(Action<TArg> executedMethod, Func<TArg, bool> canExecuteMethod)
: base(executedMethod, canExecuteMethod) { }
public DepedencyCommand(TViewModel viewModelInstance, Action<TArg> executedMethod, Func<TArg, bool> canExecuteMethod, Expression<Func<TViewModel, object>> propertiesToWatch)
: base(executedMethod, canExecuteMethod)
{
_propertiesToWatch = _RegisterPropertiesWatcher(propertiesToWatch);
viewModelInstance.PropertyChanged += PropertyChangedHandler;
}
/// <summary>
/// handler that, everytime a monitored property changes, calls the RaiseCanExecuteChanged of the command
/// </summary>
/// <param name="sender"></param>
/// <param name="e"></param>
private void PropertyChangedHandler(object sender, PropertyChangedEventArgs e)
{
if (_propertiesToWatch.Contains(e.PropertyName))
{
this.OnCanExecuteChanged();
}
}
/// <summary>
/// giving an expression that identify a propriety or a list of properties, return the property names obtained from the expression
/// Examples on selector usage
/// proprietà singola:
/// entity => entity.PropertyName
/// proprietà multiple
/// entity => new { entity.PropertyName1, entity.PropertyName2 }
/// </summary>
/// <param name="selector"></param>
/// <returns></returns>
private static List<string> _RegisterPropertiesWatcher(Expression<Func<TViewModel, object>> selector)
{
var properties = new List<string>();
LambdaExpression lambda = selector;
if (lambda.Body is MemberExpression)
{
var memberExpression = (MemberExpression)lambda.Body;
properties.Add(memberExpression.Member.Name);
}
else if (lambda.Body is UnaryExpression)
{
var unaryExpression = (UnaryExpression)lambda.Body;
properties.Add(((MemberExpression)unaryExpression.Operand).Member.Name);
}
else if (lambda.Body.NodeType == ExpressionType.New)
{
var newExp = (NewExpression)lambda.Body;
foreach (var argument in newExp.Arguments)
{
if (argument is MemberExpression)
{
MemberExpression mExp = (MemberExpression)argument;
properties.Add(mExp.Member.Name);
}
else
throw new SyntaxErrorException("Syntax Error, selector has to be an expression that returns a new object containing a list of properties, e.g.: s => new { s.Property1, s.Property2 }");
}
}
else
throw new SyntaxErrorException("Syntax Error, selector has to be an expression that returns a new object containing a list of properties, e.g.: s => new { s.Property1, s.Property2 }");
return properties;
}
}
There doesn't seem to be a better solution. I know what you mean, it seems inelegant but whatever lipstick you put on it, the onus is on the objects involved in the expression to notify the command.
If your condition is based purely on other notify properties you could add your own handler to PropertyChanged, that provides a bit of abstraction.
In this case, TimerState would be a VM property. Then you can a handler to your ViewModel.PropertyChanged event. You can then inspect the property name and update your CanExecute. This is still ugly, but at least you have all the garbage in one block.

Fast performing and thread safe observable collection

ObservableCollections raise notifications for each action performed on them. Firstly they dont have bulk add or remove calls, secondly they are not thread safe.
Doesn't this make them slower? Cant we have a faster alternative? Some say ICollectionView wrapped around an ObservableCollection is fast? How true is this claim.
ObservableCollection can be fast, if it wants to. :-)
The code below is a very good example of a thread safe, faster observable collection and you can extend it further to your wish.
using System.Collections.Specialized;
public class FastObservableCollection<T> : ObservableCollection<T>
{
private readonly object locker = new object();
/// <summary>
/// This private variable holds the flag to
/// turn on and off the collection changed notification.
/// </summary>
private bool suspendCollectionChangeNotification;
/// <summary>
/// Initializes a new instance of the FastObservableCollection class.
/// </summary>
public FastObservableCollection()
: base()
{
this.suspendCollectionChangeNotification = false;
}
/// <summary>
/// This event is overriden CollectionChanged event of the observable collection.
/// </summary>
public override event NotifyCollectionChangedEventHandler CollectionChanged;
/// <summary>
/// This method adds the given generic list of items
/// as a range into current collection by casting them as type T.
/// It then notifies once after all items are added.
/// </summary>
/// <param name="items">The source collection.</param>
public void AddItems(IList<T> items)
{
lock(locker)
{
this.SuspendCollectionChangeNotification();
foreach (var i in items)
{
InsertItem(Count, i);
}
this.NotifyChanges();
}
}
/// <summary>
/// Raises collection change event.
/// </summary>
public void NotifyChanges()
{
this.ResumeCollectionChangeNotification();
var arg
= new NotifyCollectionChangedEventArgs
(NotifyCollectionChangedAction.Reset);
this.OnCollectionChanged(arg);
}
/// <summary>
/// This method removes the given generic list of items as a range
/// into current collection by casting them as type T.
/// It then notifies once after all items are removed.
/// </summary>
/// <param name="items">The source collection.</param>
public void RemoveItems(IList<T> items)
{
lock(locker)
{
this.SuspendCollectionChangeNotification();
foreach (var i in items)
{
Remove(i);
}
this.NotifyChanges();
}
}
/// <summary>
/// Resumes collection changed notification.
/// </summary>
public void ResumeCollectionChangeNotification()
{
this.suspendCollectionChangeNotification = false;
}
/// <summary>
/// Suspends collection changed notification.
/// </summary>
public void SuspendCollectionChangeNotification()
{
this.suspendCollectionChangeNotification = true;
}
/// <summary>
/// This collection changed event performs thread safe event raising.
/// </summary>
/// <param name="e">The event argument.</param>
protected override void OnCollectionChanged(NotifyCollectionChangedEventArgs e)
{
// Recommended is to avoid reentry
// in collection changed event while collection
// is getting changed on other thread.
using (BlockReentrancy())
{
if (!this.suspendCollectionChangeNotification)
{
NotifyCollectionChangedEventHandler eventHandler =
this.CollectionChanged;
if (eventHandler == null)
{
return;
}
// Walk thru invocation list.
Delegate[] delegates = eventHandler.GetInvocationList();
foreach
(NotifyCollectionChangedEventHandler handler in delegates)
{
// If the subscriber is a DispatcherObject and different thread.
DispatcherObject dispatcherObject
= handler.Target as DispatcherObject;
if (dispatcherObject != null
&& !dispatcherObject.CheckAccess())
{
// Invoke handler in the target dispatcher's thread...
// asynchronously for better responsiveness.
dispatcherObject.Dispatcher.BeginInvoke
(DispatcherPriority.DataBind, handler, this, e);
}
else
{
// Execute handler as is.
handler(this, e);
}
}
}
}
}
}
Also ICollectionView that sits above the ObservableCollection is actively aware of the changes and performs filtering, grouping, sorting relatively fast as compared to any other source list.
Again observable collections may not be a perfect answer for faster data updates but they do their job pretty well.
Here is a compilation of some solutions which I made. The idea of collection changed invokation taken from first answer.
Also seems that "Reset" operation should be synchronous with main thread otherwise strange things happen to CollectionView and CollectionViewSource.
I think that's because on "Reset" handler tries to read the collection contents immediately and they should be already in place.
If you do "Reset" async and than immediately add some items also async than newly added items might be added twice.
public interface IObservableList<T> : IList<T>, INotifyCollectionChanged
{
}
public class ObservableList<T> : IObservableList<T>
{
private IList<T> collection = new List<T>();
public event NotifyCollectionChangedEventHandler CollectionChanged;
private ReaderWriterLock sync = new ReaderWriterLock();
protected virtual void OnCollectionChanged(NotifyCollectionChangedEventArgs args)
{
if (CollectionChanged == null)
return;
foreach (NotifyCollectionChangedEventHandler handler in CollectionChanged.GetInvocationList())
{
// If the subscriber is a DispatcherObject and different thread.
var dispatcherObject = handler.Target as DispatcherObject;
if (dispatcherObject != null && !dispatcherObject.CheckAccess())
{
if ( args.Action == NotifyCollectionChangedAction.Reset )
dispatcherObject.Dispatcher.Invoke
(DispatcherPriority.DataBind, handler, this, args);
else
// Invoke handler in the target dispatcher's thread...
// asynchronously for better responsiveness.
dispatcherObject.Dispatcher.BeginInvoke
(DispatcherPriority.DataBind, handler, this, args);
}
else
{
// Execute handler as is.
handler(this, args);
}
}
}
public ObservableList()
{
}
public void Add(T item)
{
sync.AcquireWriterLock(Timeout.Infinite);
try
{
collection.Add(item);
OnCollectionChanged(
new NotifyCollectionChangedEventArgs(
NotifyCollectionChangedAction.Add, item));
}
finally
{
sync.ReleaseWriterLock();
}
}
public void Clear()
{
sync.AcquireWriterLock(Timeout.Infinite);
try
{
collection.Clear();
OnCollectionChanged(
new NotifyCollectionChangedEventArgs(
NotifyCollectionChangedAction.Reset));
}
finally
{
sync.ReleaseWriterLock();
}
}
public bool Contains(T item)
{
sync.AcquireReaderLock(Timeout.Infinite);
try
{
var result = collection.Contains(item);
return result;
}
finally
{
sync.ReleaseReaderLock();
}
}
public void CopyTo(T[] array, int arrayIndex)
{
sync.AcquireWriterLock(Timeout.Infinite);
try
{
collection.CopyTo(array, arrayIndex);
}
finally
{
sync.ReleaseWriterLock();
}
}
public int Count
{
get
{
sync.AcquireReaderLock(Timeout.Infinite);
try
{
return collection.Count;
}
finally
{
sync.ReleaseReaderLock();
}
}
}
public bool IsReadOnly
{
get { return collection.IsReadOnly; }
}
public bool Remove(T item)
{
sync.AcquireWriterLock(Timeout.Infinite);
try
{
var index = collection.IndexOf(item);
if (index == -1)
return false;
var result = collection.Remove(item);
if (result)
OnCollectionChanged(new NotifyCollectionChangedEventArgs(NotifyCollectionChangedAction.Remove, item, index));
return result;
}
finally
{
sync.ReleaseWriterLock();
}
}
public IEnumerator<T> GetEnumerator()
{
return collection.GetEnumerator();
}
System.Collections.IEnumerator System.Collections.IEnumerable.GetEnumerator()
{
return collection.GetEnumerator();
}
public int IndexOf(T item)
{
sync.AcquireReaderLock(Timeout.Infinite);
try
{
var result = collection.IndexOf(item);
return result;
}
finally
{
sync.ReleaseReaderLock();
}
}
public void Insert(int index, T item)
{
sync.AcquireWriterLock(Timeout.Infinite);
try
{
collection.Insert(index, item);
OnCollectionChanged(new NotifyCollectionChangedEventArgs(NotifyCollectionChangedAction.Add, item, index));
}
finally
{
sync.ReleaseWriterLock();
}
}
public void RemoveAt(int index)
{
sync.AcquireWriterLock(Timeout.Infinite);
try
{
if (collection.Count == 0 || collection.Count <= index)
return;
var item = collection[index];
collection.RemoveAt(index);
OnCollectionChanged(
new NotifyCollectionChangedEventArgs(
NotifyCollectionChangedAction.Remove, item, index));
}
finally
{
sync.ReleaseWriterLock();
}
}
public T this[int index]
{
get
{
sync.AcquireReaderLock(Timeout.Infinite);
try
{
var result = collection[index];
return result;
}
finally
{
sync.ReleaseReaderLock();
}
}
set
{
sync.AcquireWriterLock(Timeout.Infinite);
try
{
if (collection.Count == 0 || collection.Count <= index)
return;
var item = collection[index];
collection[index] = value;
OnCollectionChanged(
new NotifyCollectionChangedEventArgs(
NotifyCollectionChangedAction.Replace, value, item, index));
}
finally
{
sync.ReleaseWriterLock();
}
}
}
}
I can't add comments because I'm not cool enough yet, but sharing this issue I ran into is probably worth posting even though it's not really an answer. I kept getting an "Index was out of range" exception using this FastObservableCollection, because of the BeginInvoke. Apparently changes being notified can be undone before the handler is called, so to fix this I passed the following as the fourth parameter for the BeginInvoke called from the OnCollectionChanged method (as opposed to using the event args one):
dispatcherObject.Dispatcher.BeginInvoke
(DispatcherPriority.DataBind, handler, this, new NotifyCollectionChangedEventArgs(NotifyCollectionChangedAction.Reset));
Instead of this:
dispatcherObject.Dispatcher.BeginInvoke
(DispatcherPriority.DataBind, handler, this, e);
This fixed the "Index was out of range" issue I was running into. Here's a more detailed explaination / code snpipet: Where do I get a thread-safe CollectionView?
An example where is created a synchronized Observable list:
newSeries = new XYChart.Series<>();
ObservableList<XYChart.Data<Number, Number>> listaSerie;
listaSerie = FXCollections.synchronizedObservableList(FXCollections.observableList(new ArrayList<XYChart.Data<Number, Number>>()));
newSeries.setData(listaSerie);

CommandManager.InvalidateRequerySuggested() isn't fast enough. What can I do?

Short Version
Calls to CommandManager.InvalidateRequerySuggested() take far longer to take effect than I would like (1-2 second delay before UI controls become disabled).
Long Version
I have a system where I submit tasks to a background-thread based task processor. This submit happens on the WPF UI thread.
When this submit happens, the object that manages my background thread does two things:
It raises a "busy" event (still on the UI thread) that several view models respond to; when they receive this event, they set an IsEnabled flag on themselves to false. Controls in my views, which are databound to this property, are immediately grayed out, which is what I would expect.
It informs my WPF ICommand objects that they should not be allowed to execute (again, still on the UI thread). Because there is nothing like INotifyPropertyChanged for ICommand objects, I am forced to call CommandManager.InvalidateRequerySuggested() to force WPF to reconsider all of my command objects' CanExecute states (yes, I actually do need to do this: otherwise, none of these controls become disabled). Unlike item 1, though, it takes a significantly longer time for my buttons/menu items/etc that are using ICommand objects to visually change to a disabled state than it does for the UI controls that have their IsEnabled property manually set.
The problem is, from a UX point of view, this looks awful; half of my controls are immediately grayed out (because their IsEnabled property is set to false), and then a full 1-2 seconds later, the other half of my controls follow suit (because their CanExecute methods are finally re-evaluated).
So, part 1 of my question:
As silly as it sounds to ask, is there a way I can make CommandManager.InvalidateRequerySuggested() do it's job faster? I suspect that there isn't.
Fair enough, part 2 of my question:
How can I work around this? I'd prefer all of my controls be disabled at the same time. It just looks unprofessional and awkward otherwise. Any ideas? :-)
CommandManager.InvalidateRequerySuggested() tries to validate all commands, which is totally ineffective (and in your case slow) - on every change, you are asking every command to recheck its CanExecute()!
You'd need the command to know on which objects and properties is its CanExecute dependent, and suggest requery only when they change. That way, if you change a property of an object, only commands that depend on it will change their state.
This is how I solved the problem, but at first, a teaser:
// in ViewModel's constructor - add a code to public ICommand:
this.DoStuffWithParameterCommand = new DelegateCommand<object>(
parameter =>
{
//do work with parameter (remember to check against null)
},
parameter =>
{
//can this command execute? return true or false
}
)
.ListenOn(whichObject, n => n.ObjectProperty /*type safe!*/, this.Dispatcher /*we need to pass UI dispatcher here*/)
.ListenOn(anotherObject, n => n.AnotherObjectProperty, this.Dispatcher); // chain calling!
The command is listening on NotifyPropertyChanged events from object that affect whether it can execute, and invokes the check only when a requery is needed.
Now, a lot of code (part of our in-house framework) to do this:
I use DelegateCommand from Prism, that looks like this:
/// <summary>
/// This class allows delegating the commanding logic to methods passed as parameters,
/// and enables a View to bind commands to objects that are not part of the element tree.
/// </summary>
public class DelegateCommand : ICommand
{
#region Constructors
/// <summary>
/// Constructor
/// </summary>
public DelegateCommand(Action executeMethod)
: this(executeMethod, null, false)
{
}
/// <summary>
/// Constructor
/// </summary>
public DelegateCommand(Action executeMethod, Func<bool> canExecuteMethod)
: this(executeMethod, canExecuteMethod, false)
{
}
/// <summary>
/// Constructor
/// </summary>
public DelegateCommand(Action executeMethod, Func<bool> canExecuteMethod, bool isAutomaticRequeryDisabled)
{
if (executeMethod == null)
{
throw new ArgumentNullException("executeMethod");
}
_executeMethod = executeMethod;
_canExecuteMethod = canExecuteMethod;
_isAutomaticRequeryDisabled = isAutomaticRequeryDisabled;
this.RaiseCanExecuteChanged();
}
#endregion
#region Public Methods
/// <summary>
/// Method to determine if the command can be executed
/// </summary>
public bool CanExecute()
{
if (_canExecuteMethod != null)
{
return _canExecuteMethod();
}
return true;
}
/// <summary>
/// Execution of the command
/// </summary>
public void Execute()
{
if (_executeMethod != null)
{
_executeMethod();
}
}
/// <summary>
/// Property to enable or disable CommandManager's automatic requery on this command
/// </summary>
public bool IsAutomaticRequeryDisabled
{
get
{
return _isAutomaticRequeryDisabled;
}
set
{
if (_isAutomaticRequeryDisabled != value)
{
if (value)
{
CommandManagerHelper.RemoveHandlersFromRequerySuggested(_canExecuteChangedHandlers);
}
else
{
CommandManagerHelper.AddHandlersToRequerySuggested(_canExecuteChangedHandlers);
}
_isAutomaticRequeryDisabled = value;
}
}
}
/// <summary>
/// Raises the CanExecuteChaged event
/// </summary>
public void RaiseCanExecuteChanged()
{
OnCanExecuteChanged();
}
/// <summary>
/// Protected virtual method to raise CanExecuteChanged event
/// </summary>
protected virtual void OnCanExecuteChanged()
{
CommandManagerHelper.CallWeakReferenceHandlers(_canExecuteChangedHandlers);
}
#endregion
#region ICommand Members
/// <summary>
/// ICommand.CanExecuteChanged implementation
/// </summary>
public event EventHandler CanExecuteChanged
{
add
{
if (!_isAutomaticRequeryDisabled)
{
CommandManager.RequerySuggested += value;
}
CommandManagerHelper.AddWeakReferenceHandler(ref _canExecuteChangedHandlers, value, 2);
}
remove
{
if (!_isAutomaticRequeryDisabled)
{
CommandManager.RequerySuggested -= value;
}
CommandManagerHelper.RemoveWeakReferenceHandler(_canExecuteChangedHandlers, value);
}
}
bool ICommand.CanExecute(object parameter)
{
return CanExecute();
}
void ICommand.Execute(object parameter)
{
Execute();
}
#endregion
#region Data
private readonly Action _executeMethod = null;
private readonly Func<bool> _canExecuteMethod = null;
private bool _isAutomaticRequeryDisabled = false;
private List<WeakReference> _canExecuteChangedHandlers;
#endregion
}
/// <summary>
/// This class allows delegating the commanding logic to methods passed as parameters,
/// and enables a View to bind commands to objects that are not part of the element tree.
/// </summary>
/// <typeparam name="T">Type of the parameter passed to the delegates</typeparam>
public class DelegateCommand<T> : ICommand
{
#region Constructors
/// <summary>
/// Constructor
/// </summary>
public DelegateCommand(Action<T> executeMethod)
: this(executeMethod, null, false)
{
}
/// <summary>
/// Constructor
/// </summary>
public DelegateCommand(Action<T> executeMethod, Func<T, bool> canExecuteMethod)
: this(executeMethod, canExecuteMethod, false)
{
}
/// <summary>
/// Constructor
/// </summary>
public DelegateCommand(Action<T> executeMethod, Func<T, bool> canExecuteMethod, bool isAutomaticRequeryDisabled)
{
if (executeMethod == null)
{
throw new ArgumentNullException("executeMethod");
}
_executeMethod = executeMethod;
_canExecuteMethod = canExecuteMethod;
_isAutomaticRequeryDisabled = isAutomaticRequeryDisabled;
}
#endregion
#region Public Methods
/// <summary>
/// Method to determine if the command can be executed
/// </summary>
public bool CanExecute(T parameter)
{
if (_canExecuteMethod != null)
{
return _canExecuteMethod(parameter);
}
return true;
}
/// <summary>
/// Execution of the command
/// </summary>
public void Execute(T parameter)
{
if (_executeMethod != null)
{
_executeMethod(parameter);
}
}
/// <summary>
/// Raises the CanExecuteChaged event
/// </summary>
public void RaiseCanExecuteChanged()
{
OnCanExecuteChanged();
}
/// <summary>
/// Protected virtual method to raise CanExecuteChanged event
/// </summary>
protected virtual void OnCanExecuteChanged()
{
CommandManagerHelper.CallWeakReferenceHandlers(_canExecuteChangedHandlers);
}
/// <summary>
/// Property to enable or disable CommandManager's automatic requery on this command
/// </summary>
public bool IsAutomaticRequeryDisabled
{
get
{
return _isAutomaticRequeryDisabled;
}
set
{
if (_isAutomaticRequeryDisabled != value)
{
if (value)
{
CommandManagerHelper.RemoveHandlersFromRequerySuggested(_canExecuteChangedHandlers);
}
else
{
CommandManagerHelper.AddHandlersToRequerySuggested(_canExecuteChangedHandlers);
}
_isAutomaticRequeryDisabled = value;
}
}
}
#endregion
#region ICommand Members
/// <summary>
/// ICommand.CanExecuteChanged implementation
/// </summary>
public event EventHandler CanExecuteChanged
{
add
{
if (!_isAutomaticRequeryDisabled)
{
CommandManager.RequerySuggested += value;
}
CommandManagerHelper.AddWeakReferenceHandler(ref _canExecuteChangedHandlers, value, 2);
}
remove
{
if (!_isAutomaticRequeryDisabled)
{
CommandManager.RequerySuggested -= value;
}
CommandManagerHelper.RemoveWeakReferenceHandler(_canExecuteChangedHandlers, value);
}
}
bool ICommand.CanExecute(object parameter)
{
// if T is of value type and the parameter is not
// set yet, then return false if CanExecute delegate
// exists, else return true
if (parameter == null &&
typeof(T).IsValueType)
{
return (_canExecuteMethod == null);
}
return CanExecute((T)parameter);
}
void ICommand.Execute(object parameter)
{
Execute((T)parameter);
}
#endregion
#region Data
private readonly Action<T> _executeMethod = null;
private readonly Func<T, bool> _canExecuteMethod = null;
private bool _isAutomaticRequeryDisabled = false;
private List<WeakReference> _canExecuteChangedHandlers;
#endregion
}
/// <summary>
/// This class contains methods for the CommandManager that help avoid memory leaks by
/// using weak references.
/// </summary>
internal class CommandManagerHelper
{
internal static void CallWeakReferenceHandlers(List<WeakReference> handlers)
{
if (handlers != null)
{
// Take a snapshot of the handlers before we call out to them since the handlers
// could cause the array to me modified while we are reading it.
EventHandler[] callees = new EventHandler[handlers.Count];
int count = 0;
for (int i = handlers.Count - 1; i >= 0; i--)
{
WeakReference reference = handlers[i];
EventHandler handler = reference.Target as EventHandler;
if (handler == null)
{
// Clean up old handlers that have been collected
handlers.RemoveAt(i);
}
else
{
callees[count] = handler;
count++;
}
}
// Call the handlers that we snapshotted
for (int i = 0; i < count; i++)
{
EventHandler handler = callees[i];
handler(null, EventArgs.Empty);
}
}
}
internal static void AddHandlersToRequerySuggested(List<WeakReference> handlers)
{
if (handlers != null)
{
foreach (WeakReference handlerRef in handlers)
{
EventHandler handler = handlerRef.Target as EventHandler;
if (handler != null)
{
CommandManager.RequerySuggested += handler;
}
}
}
}
internal static void RemoveHandlersFromRequerySuggested(List<WeakReference> handlers)
{
if (handlers != null)
{
foreach (WeakReference handlerRef in handlers)
{
EventHandler handler = handlerRef.Target as EventHandler;
if (handler != null)
{
CommandManager.RequerySuggested -= handler;
}
}
}
}
internal static void AddWeakReferenceHandler(ref List<WeakReference> handlers, EventHandler handler)
{
AddWeakReferenceHandler(ref handlers, handler, -1);
}
internal static void AddWeakReferenceHandler(ref List<WeakReference> handlers, EventHandler handler, int defaultListSize)
{
if (handlers == null)
{
handlers = (defaultListSize > 0 ? new List<WeakReference>(defaultListSize) : new List<WeakReference>());
}
handlers.Add(new WeakReference(handler));
}
internal static void RemoveWeakReferenceHandler(List<WeakReference> handlers, EventHandler handler)
{
if (handlers != null)
{
for (int i = handlers.Count - 1; i >= 0; i--)
{
WeakReference reference = handlers[i];
EventHandler existingHandler = reference.Target as EventHandler;
if ((existingHandler == null) || (existingHandler == handler))
{
// Clean up old handlers that have been collected
// in addition to the handler that is to be removed.
handlers.RemoveAt(i);
}
}
}
}
}
I have then written a ListenOn extension method, that 'binds' the command to a property, and invokes its RaiseCanExecuteChanged:
public static class DelegateCommandExtensions
{
/// <summary>
/// Makes DelegateCommnand listen on PropertyChanged events of some object,
/// so that DelegateCommnand can update its IsEnabled property.
/// </summary>
public static DelegateCommand ListenOn<ObservedType, PropertyType>
(this DelegateCommand delegateCommand,
ObservedType observedObject,
Expression<Func<ObservedType, PropertyType>> propertyExpression,
Dispatcher dispatcher)
where ObservedType : INotifyPropertyChanged
{
//string propertyName = observedObject.GetPropertyName(propertyExpression);
string propertyName = NotifyPropertyChangedBaseExtensions.GetPropertyName(propertyExpression);
observedObject.PropertyChanged += (sender, e) =>
{
if (e.PropertyName == propertyName)
{
if (dispatcher != null)
{
ThreadTools.RunInDispatcher(dispatcher, delegateCommand.RaiseCanExecuteChanged);
}
else
{
delegateCommand.RaiseCanExecuteChanged();
}
}
};
return delegateCommand; //chain calling
}
/// <summary>
/// Makes DelegateCommnand listen on PropertyChanged events of some object,
/// so that DelegateCommnand can update its IsEnabled property.
/// </summary>
public static DelegateCommand<T> ListenOn<T, ObservedType, PropertyType>
(this DelegateCommand<T> delegateCommand,
ObservedType observedObject,
Expression<Func<ObservedType, PropertyType>> propertyExpression,
Dispatcher dispatcher)
where ObservedType : INotifyPropertyChanged
{
//string propertyName = observedObject.GetPropertyName(propertyExpression);
string propertyName = NotifyPropertyChangedBaseExtensions.GetPropertyName(propertyExpression);
observedObject.PropertyChanged += (object sender, PropertyChangedEventArgs e) =>
{
if (e.PropertyName == propertyName)
{
if (dispatcher != null)
{
ThreadTools.RunInDispatcher(dispatcher, delegateCommand.RaiseCanExecuteChanged);
}
else
{
delegateCommand.RaiseCanExecuteChanged();
}
}
};
return delegateCommand; //chain calling
}
}
You then need the following extension to NotifyPropertyChanged
/// <summary>
/// <see cref="http://dotnet.dzone.com/news/silverlightwpf-implementing"/>
/// </summary>
public static class NotifyPropertyChangedBaseExtensions
{
/// <summary>
/// Raises PropertyChanged event.
/// To use: call the extension method with this: this.OnPropertyChanged(n => n.Title);
/// </summary>
/// <typeparam name="T">Property owner</typeparam>
/// <typeparam name="TProperty">Type of property</typeparam>
/// <param name="observableBase"></param>
/// <param name="expression">Property expression like 'n => n.Property'</param>
public static void OnPropertyChanged<T, TProperty>(this T observableBase, Expression<Func<T, TProperty>> expression) where T : INotifyPropertyChangedWithRaise
{
observableBase.OnPropertyChanged(GetPropertyName<T, TProperty>(expression));
}
public static string GetPropertyName<T, TProperty>(Expression<Func<T, TProperty>> expression) where T : INotifyPropertyChanged
{
if (expression == null)
throw new ArgumentNullException("expression");
var lambda = expression as LambdaExpression;
MemberExpression memberExpression;
if (lambda.Body is UnaryExpression)
{
var unaryExpression = lambda.Body as UnaryExpression;
memberExpression = unaryExpression.Operand as MemberExpression;
}
else
{
memberExpression = lambda.Body as MemberExpression;
}
if (memberExpression == null)
throw new ArgumentException("Please provide a lambda expression like 'n => n.PropertyName'");
MemberInfo memberInfo = memberExpression.Member;
if (String.IsNullOrEmpty(memberInfo.Name))
throw new ArgumentException("'expression' did not provide a property name.");
return memberInfo.Name;
}
}
where INotifyPropertyChangedWithRaise is this (it estabilishes standard interface for raising NotifyPropertyChanged events):
public interface INotifyPropertyChangedWithRaise : INotifyPropertyChanged
{
void OnPropertyChanged(string propertyName);
}
Last piece of puzzle is this:
public class ThreadTools
{
public static void RunInDispatcher(Dispatcher dispatcher, Action action)
{
RunInDispatcher(dispatcher, DispatcherPriority.Normal, action);
}
public static void RunInDispatcher(Dispatcher dispatcher, DispatcherPriority priority, Action action)
{
if (action == null) { return; }
if (dispatcher.CheckAccess())
{
// we are already on thread associated with the dispatcher -> just call action
try
{
action();
}
catch (Exception ex)
{
//Log error here!
}
}
else
{
// we are on different thread, invoke action on dispatcher's thread
dispatcher.BeginInvoke(
priority,
(Action)(
() =>
{
try
{
action();
}
catch (Exception ex)
{
//Log error here!
}
})
);
}
}
}
This solution is a reduced version of the solution proposed by Tomáš Kafka(thanks to Tomas for describing his solution in detail)in this thread.
In Tomas's solution he had
1) DelegateCommand
2) CommandManagerHelper
3) DelegateCommandExtensions
4) NotifyPropertyChangedBaseExtensions
5) INotifyPropertyChangedWithRaise
6) ThreadTools
This solution has
1) DelegateCommand
2) DelegateCommandExtensions method and NotifyPropertyChangedBaseExtensions method in Delegate Command itself.
Note Since our wpf application follows MVVM pattern and we handle commands at viewmodel level which executes in UI thread we don't need to get the reference to UI disptacher.
using System;
using System.Collections.Generic;
using System.ComponentModel;
using System.Linq.Expressions;
using System.Reflection;
using System.Windows.Input;
namespace ExampleForDelegateCommand
{
public class DelegateCommand : ICommand
{
public Predicate<object> CanExecuteDelegate { get; set; }
private List<INotifyPropertyChanged> propertiesToListenTo;
private List<WeakReference> ControlEvent;
public DelegateCommand()
{
ControlEvent= new List<WeakReference>();
}
public List<INotifyPropertyChanged> PropertiesToListenTo
{
get { return propertiesToListenTo; }
set
{
propertiesToListenTo = value;
}
}
private Action<object> executeDelegate;
public Action<object> ExecuteDelegate
{
get { return executeDelegate; }
set
{
executeDelegate = value;
ListenForNotificationFrom((INotifyPropertyChanged)executeDelegate.Target);
}
}
public static ICommand Create(Action<object> exec)
{
return new SimpleCommand { ExecuteDelegate = exec };
}
#region ICommand Members
public bool CanExecute(object parameter)
{
if (CanExecuteDelegate != null)
return CanExecuteDelegate(parameter);
return true; // if there is no can execute default to true
}
public event EventHandler CanExecuteChanged
{
add
{
CommandManager.RequerySuggested += value;
ControlEvent.Add(new WeakReference(value));
}
remove
{
CommandManager.RequerySuggested -= value;
ControlEvent.Remove(ControlEvent.Find(r => ((EventHandler) r.Target) == value));
}
}
public void Execute(object parameter)
{
if (ExecuteDelegate != null)
ExecuteDelegate(parameter);
}
#endregion
public void RaiseCanExecuteChanged()
{
if (ControlEvent != null && ControlEvent.Count > 0)
{
ControlEvent.ForEach(ce =>
{
if(ce.Target!=null)
((EventHandler) (ce.Target)).Invoke(null, EventArgs.Empty);
});
}
}
public DelegateCommand ListenOn<TObservedType, TPropertyType>(TObservedType viewModel, Expression<Func<TObservedType, TPropertyType>> propertyExpression) where TObservedType : INotifyPropertyChanged
{
string propertyName = GetPropertyName(propertyExpression);
viewModel.PropertyChanged += (PropertyChangedEventHandler)((sender, e) =>
{
if (e.PropertyName == propertyName) RaiseCanExecuteChanged();
});
return this;
}
public void ListenForNotificationFrom<TObservedType>(TObservedType viewModel) where TObservedType : INotifyPropertyChanged
{
viewModel.PropertyChanged += (PropertyChangedEventHandler)((sender, e) =>
{
RaiseCanExecuteChanged();
});
}
private string GetPropertyName<T, TProperty>(Expression<Func<T, TProperty>> expression) where T : INotifyPropertyChanged
{
var lambda = expression as LambdaExpression;
MemberInfo memberInfo = GetmemberExpression(lambda).Member;
return memberInfo.Name;
}
private MemberExpression GetmemberExpression(LambdaExpression lambda)
{
MemberExpression memberExpression;
if (lambda.Body is UnaryExpression)
{
var unaryExpression = lambda.Body as UnaryExpression;
memberExpression = unaryExpression.Operand as MemberExpression;
}
else
memberExpression = lambda.Body as MemberExpression;
return memberExpression;
}
}}
Explanation of the solution:
Normally when we bind a UI element(Button)to the ICommand implementation the WPF Button registers for a Event "CanExecuteChanged" in ICommand implementation .If your Icommand implementation for "CanExecuteChanged" hook to the CommandManager's RequesySuggest event(read this article http://joshsmithonwpf.wordpress.com/2008/06/17/allowing-commandmanager-to-query-your-icommand-objects/) then when ever CommandManager detects conditions that might change the ability of a command to execute(changes like Focus shifts and some keyboard events) , CommandManager’s RequerySuggested event occurs which in turn will cause Button'e delegate to be called since we hooked the buttos's delgate to CommandManager’s RequerySuggested in the implementation of "CanExecuteChanged" in our DelegateCommand .
But the problem is that ComandManager is not able to always detect the changes. Hence the solution it to raise "CanExecuteChanged" when our command implementation(DelegateCommand) detects there is a change.Normally when we declare the delagate for ICommand's CanExecute in our viewmodel we bind to properties declared in our viewmodel and our ICommand implementation can listen for "propertychanged" events on these properties. Thats what the "ListenForNotificationFrom" method of the DelegateCommand does. In case the client code does not register for specific property changes the DelegateCommand by defaults listens to any property change on the view model where command is declared and defined.
"ControlEvent" in DelegateCommand which is list of EventHandler that stores the Button's
"CanExecuteChange EventHandler" is declared as weak reference to avoid memory leaks.
How will ViewModel use this DelegateCommand
There are 2 ways to use it.
(the second usage is more specific to the properties that you want the Command to listen to.
delegateCommand = new DelegateCommand
{
ExecuteDelegate = Search,
CanExecuteDelegate = (r) => !IsBusy
};
anotherDelegateCommand = new DelegateCommand
{
ExecuteDelegate = SearchOne,
CanExecuteDelegate = (r) => !IsBusyOne
}.ListenOn(this, n => n.IsBusyOne);
A detailed ViewModel
public class ExampleViewModel
{
public SearchViewModelBase()
{
delegateCommand = new DelegateCommand
{
ExecuteDelegate = Search,
CanExecuteDelegate = (r) => !IsBusy
};
anotherDelegateCommand = new DelegateCommand
{
ExecuteDelegate = SearchOne,
CanExecuteDelegate = (r) => !IsBusyOne
}.ListenOn(this, n => n.IsBusyOne);
}
private bool isBusy;
public virtual bool IsBusy
{
get { return isBusy; }
set
{
if (isBusy == value) return;
isBusy = value;
NotifyPropertyChanged(MethodBase.GetCurrentMethod());
}
}
private bool isBusyOne;
public virtual bool IsBusyOne
{
get { return isBusyOne; }
set
{
if (isBusyOne == value) return;
isBusyOne = value;
NotifyPropertyChanged(MethodBase.GetCurrentMethod());
}
}
private void Search(object obj)
{
IsBusy = true;
new SearchService().Search(Callback);
}
public void Callback(ServiceResponse response)
{
IsBusy = false;
}
private void Search(object obj)
{
IsBusyOne = true;
new SearchService().Search(CallbackOne);
}
public void CallbackOne(ServiceResponse response)
{
IsBusyOne = false;
}
private void NotifyPropertyChanged(string propertyName)
{
if (PropertyChanged != null)
{
PropertyChanged(this, new PropertyChangedEventArgs(propertyName));
}
}
private void NotifyPropertyChanged(MethodBase methodBase)
{
string methodName = methodBase.Name;
if (!methodName.StartsWith("set_"))
{
var ex = new ArgumentException("MethodBase must refer to a Property Setter method.");
throw ex;
}
NotifyPropertyChanged(methodName.Substring(4));
}
}
Tomas has a nice solution, but pls note there's a serious bug in that the CanExecute will not always fire when bound to a Button due to this :
// Call the handlers that we snapshotted
for (int i = 0; i < count; i++)
{
EventHandler handler = callees[i];
handler(null, EventArgs.Empty);
}
The 'null' parameter passed in causes issues with the CanExecuteChangedEventManager (used by the WPF Button class to listen to changes on any Command bound to it). Specifically, the CanExecuteChangedEventManager maintains a collection of weak events that need to be invoked to determine if the command Can-Execute() but this collection is keyed by the 'sender'.
The fix is simple and works for me - change the signature to
internal static void CallWeakReferenceHandlers(ICommand sender, List<WeakReference> handlers)
{
....
handler(sender, EventArgs.Empty);
}
Sorry I haven't described it too well - in a bit of a rush to catch up with my dev now after taking a few hours to figure this out !
I would suggest looking into ReactiveUI and specifically at the ICommand implementation it provides, ReactiveCommand. It uses a different approach than DelegateCommand/RelayCommand which are implemented with delegates for CanExecute that must be actively evaluated. ReactiveCommand's value for CanExecute is pushed using IObservables.
is there a way I can make CommandManager.InvalidateRequerySuggested() do it's job faster?
Yes, there is way to make it work faster!
Implement Command to keep / cache CanExecuteState in a boolean variable.
Implement RaiseCanExecuteChanged method to recalculate CanExecuteState and if it really changed to raise CanExecuteChanged event.
Implement CanExecute method to simply return CanExecuteState.
When InvalidateRequerySuggested method is invoked Command subscribers will only read CanExecuteState variable by invoking CanExecute method and check if it changed or not. That's almost zero overhead. All Commands will be disabled / enabled almost the same time.
All work will be done in RaiseCanExecuteChanged method that will be called only once for a Command and only for a limited set of Commands.
Try writing your own binding that calls your RaiseCanExecuteChanged() within converts? it is easier
Just to clarify:
You want to fire an update of CanExecute when Command property changed
Create your own binding class that detect changes in the Command property and then calls RaiseCanExecuteChanged()
Use this binding in CommandParameter
Worked for me.

When Clearing an ObservableCollection, There are No Items in e.OldItems

I have something here that is really catching me off guard.
I have an ObservableCollection of T that is filled with items. I also have an event handler attached to the CollectionChanged event.
When you Clear the collection it causes an CollectionChanged event with e.Action set to NotifyCollectionChangedAction.Reset. Ok, that's normal. But what is weird is that neither e.OldItems or e.NewItems has anything in it. I would expect e.OldItems to be filled with all items that were removed from the collection.
Has anyone else seen this? And if so, how have they gotten around it?
Some background: I am using the CollectionChanged event to attach and detach from another event and thus if I don't get any items in e.OldItems ... I won't be able to detach from that event.
CLARIFICATION:
I do know that the documentation doesn't outright state that it has to behave this way. But for every other action, it is notifying me of what it has done. So, my assumption is that it would tell me ... in the case of Clear/Reset as well.
Below is the sample code if you wish to reproduce it yourself. First off the xaml:
<Window
x:Class="ObservableCollection.Window1"
xmlns="http://schemas.microsoft.com/winfx/2006/xaml/presentation"
xmlns:x="http://schemas.microsoft.com/winfx/2006/xaml"
Title="Window1"
Height="300"
Width="300"
>
<StackPanel>
<Button x:Name="addButton" Content="Add" Width="100" Height="25" Margin="10" Click="addButton_Click"/>
<Button x:Name="moveButton" Content="Move" Width="100" Height="25" Margin="10" Click="moveButton_Click"/>
<Button x:Name="removeButton" Content="Remove" Width="100" Height="25" Margin="10" Click="removeButton_Click"/>
<Button x:Name="replaceButton" Content="Replace" Width="100" Height="25" Margin="10" Click="replaceButton_Click"/>
<Button x:Name="resetButton" Content="Reset" Width="100" Height="25" Margin="10" Click="resetButton_Click"/>
</StackPanel>
</Window>
Next, the code behind:
using System;
using System.Collections.Generic;
using System.Linq;
using System.Text;
using System.Windows;
using System.Windows.Controls;
using System.Windows.Data;
using System.Windows.Documents;
using System.Windows.Input;
using System.Windows.Media;
using System.Windows.Media.Imaging;
using System.Windows.Navigation;
using System.Windows.Shapes;
using System.Collections.ObjectModel;
namespace ObservableCollection
{
/// <summary>
/// Interaction logic for Window1.xaml
/// </summary>
public partial class Window1 : Window
{
public Window1()
{
InitializeComponent();
_integerObservableCollection.CollectionChanged += new System.Collections.Specialized.NotifyCollectionChangedEventHandler(_integerObservableCollection_CollectionChanged);
}
private void _integerObservableCollection_CollectionChanged(object sender, System.Collections.Specialized.NotifyCollectionChangedEventArgs e)
{
switch (e.Action)
{
case System.Collections.Specialized.NotifyCollectionChangedAction.Add:
break;
case System.Collections.Specialized.NotifyCollectionChangedAction.Move:
break;
case System.Collections.Specialized.NotifyCollectionChangedAction.Remove:
break;
case System.Collections.Specialized.NotifyCollectionChangedAction.Replace:
break;
case System.Collections.Specialized.NotifyCollectionChangedAction.Reset:
break;
default:
break;
}
}
private void addButton_Click(object sender, RoutedEventArgs e)
{
_integerObservableCollection.Add(25);
}
private void moveButton_Click(object sender, RoutedEventArgs e)
{
_integerObservableCollection.Move(0, 19);
}
private void removeButton_Click(object sender, RoutedEventArgs e)
{
_integerObservableCollection.RemoveAt(0);
}
private void replaceButton_Click(object sender, RoutedEventArgs e)
{
_integerObservableCollection[0] = 50;
}
private void resetButton_Click(object sender, RoutedEventArgs e)
{
_integerObservableCollection.Clear();
}
private ObservableCollection<int> _integerObservableCollection = new ObservableCollection<int> { 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19 };
}
}
It doesn't claim to include the old items, because Reset doesn't mean that the list has been cleared
It means that some dramatic thing has taken place, and the cost of working out the add/removes would most likely exceed the cost of just re-scanning the list from scratch... so that's what you should do.
MSDN suggests an example of the entire collection being re-sorted as a candidate for reset.
To reiterate. Reset doesn't mean clear, it means Your assumptions about the list are now invalid. Treat it as if it's an entirely new list. Clear happens to be one instance of this, but there could well be others.
Some examples:
I've had a list like this with a lot of items in it, and it has been databound to a WPF ListView to display on-screen.
If you clear the list and raise the .Reset event, the performance is pretty much instant, but if you instead raise many individual .Remove events, the performance is terrible, as WPF removes the items one by one.
I've also used .Reset in my own code to indicate that the list has been re-sorted, rather than issuing thousands of individual Move operations. As with Clear, there is a large performance hit when when raising many individual events.
We had the same issue here. The Reset action in CollectionChanged does not include the OldItems. We had a workaround: we used instead the following extension method:
public static void RemoveAll(this IList list)
{
while (list.Count > 0)
{
list.RemoveAt(list.Count - 1);
}
}
We ended up not supporting the Clear() function, and throwing a NotSupportedException in CollectionChanged event for Reset actions. The RemoveAll will trigger a Remove action in CollectionChanged event, with the proper OldItems.
Okay, I know this is a very old question but I have come up with a good solution to the issue and thought I would share.
This solution takes inspiration from a lot of the great answers here but has the following advantages:
No need to create a new class and override methods from ObservableCollection
Does not tamper with the workings of NotifyCollectionChanged (so no messing with Reset)
Does not make use of reflection
Here is the code:
public static void Clear<T>(this ObservableCollection<T> collection, Action<ObservableCollection<T>> unhookAction)
{
unhookAction.Invoke(collection);
collection.Clear();
}
This extension method simply takes an Action which will be invoked before the collection is cleared.
Another option is to replace the Reset event with a single Remove event that has all the cleared items in its OldItems property as follows:
public class ObservableCollectionNoReset<T> : ObservableCollection<T>
{
protected override void ClearItems()
{
List<T> removed = new List<T>(this);
base.ClearItems();
base.OnCollectionChanged(new NotifyCollectionChangedEventArgs(NotifyCollectionChangedAction.Remove, removed));
}
protected override void OnCollectionChanged(NotifyCollectionChangedEventArgs e)
{
if (e.Action != NotifyCollectionChangedAction.Reset)
base.OnCollectionChanged(e);
}
// Constructors omitted
...
}
Advantages:
No need to subscribe to an additional event (as required by accepted answer)
Doesn't generate an event for each object removed (some other proposed solutions result in multiple Removed events).
Subscriber only needs to check NewItems & OldItems on any event to add/remove event handlers as required.
Disadvantages:
No Reset event
Small (?) overhead creating copy of list.
???
EDIT 2012-02-23
Unfortunately, when bound to WPF list based controls, Clearing a ObservableCollectionNoReset collection with multiple elements will result in an exception "Range actions not supported".
To be used with controls with this limitation, I changed the ObservableCollectionNoReset class to:
public class ObservableCollectionNoReset<T> : ObservableCollection<T>
{
// Some CollectionChanged listeners don't support range actions.
public Boolean RangeActionsSupported { get; set; }
protected override void ClearItems()
{
if (RangeActionsSupported)
{
List<T> removed = new List<T>(this);
base.ClearItems();
base.OnCollectionChanged(new NotifyCollectionChangedEventArgs(NotifyCollectionChangedAction.Remove, removed));
}
else
{
while (Count > 0 )
base.RemoveAt(Count - 1);
}
}
protected override void OnCollectionChanged(NotifyCollectionChangedEventArgs e)
{
if (e.Action != NotifyCollectionChangedAction.Reset)
base.OnCollectionChanged(e);
}
public ObservableCollectionNoReset(Boolean rangeActionsSupported = false)
{
RangeActionsSupported = rangeActionsSupported;
}
// Additional constructors omitted.
}
This isn't as efficient when RangeActionsSupported is false (the default) because one Remove notification is generated per object in the collection
Ok, even though I still wish that ObservableCollection behaved as I wished ... the code below is what I ended up doing. Basically, I created a new collection of T called TrulyObservableCollection and overrided the ClearItems method which I then used to raise a Clearing event.
In the code that uses this TrulyObservableCollection, I use this Clearing event to loop through the items that are still in the collection at that point to do the detach on the event that I was wishing to detach from.
Hope this approach helps someone else as well.
public class TrulyObservableCollection<T> : ObservableCollection<T>
{
public event EventHandler<EventArgs> Clearing;
protected virtual void OnClearing(EventArgs e)
{
if (Clearing != null)
Clearing(this, e);
}
protected override void ClearItems()
{
OnClearing(EventArgs.Empty);
base.ClearItems();
}
}
I've found a solution that allows the user to both capitalize on the efficiency of adding or removing many items at a time while only firing one event - and satisfy the needs of UIElements to get the Action.Reset event args while all other users would like a list of elements added and removed.
This solution involves overriding the CollectionChanged event. When we go to fire this event, we can actually look at the target of each registered handler and determine their type. Since only ICollectionView classes require NotifyCollectionChangedAction.Reset args when more than one item changes, we can single them out, and give everyone else proper event args that contain the full list of items removed or added. Below is the implementation.
public class BaseObservableCollection<T> : ObservableCollection<T>
{
//Flag used to prevent OnCollectionChanged from firing during a bulk operation like Add(IEnumerable<T>) and Clear()
private bool _SuppressCollectionChanged = false;
/// Overridden so that we may manually call registered handlers and differentiate between those that do and don't require Action.Reset args.
public override event NotifyCollectionChangedEventHandler CollectionChanged;
public BaseObservableCollection() : base(){}
public BaseObservableCollection(IEnumerable<T> data) : base(data){}
#region Event Handlers
protected override void OnCollectionChanged(NotifyCollectionChangedEventArgs e)
{
if( !_SuppressCollectionChanged )
{
base.OnCollectionChanged(e);
if( CollectionChanged != null )
CollectionChanged.Invoke(this, e);
}
}
//CollectionViews raise an error when they are passed a NotifyCollectionChangedEventArgs that indicates more than
//one element has been added or removed. They prefer to receive a "Action=Reset" notification, but this is not suitable
//for applications in code, so we actually check the type we're notifying on and pass a customized event args.
protected virtual void OnCollectionChangedMultiItem(NotifyCollectionChangedEventArgs e)
{
NotifyCollectionChangedEventHandler handlers = this.CollectionChanged;
if( handlers != null )
foreach( NotifyCollectionChangedEventHandler handler in handlers.GetInvocationList() )
handler(this, !(handler.Target is ICollectionView) ? e : new NotifyCollectionChangedEventArgs(NotifyCollectionChangedAction.Reset));
}
#endregion
#region Extended Collection Methods
protected override void ClearItems()
{
if( this.Count == 0 ) return;
List<T> removed = new List<T>(this);
_SuppressCollectionChanged = true;
base.ClearItems();
_SuppressCollectionChanged = false;
OnCollectionChangedMultiItem(new NotifyCollectionChangedEventArgs(NotifyCollectionChangedAction.Remove, removed));
}
public void Add(IEnumerable<T> toAdd)
{
if( this == toAdd )
throw new Exception("Invalid operation. This would result in iterating over a collection as it is being modified.");
_SuppressCollectionChanged = true;
foreach( T item in toAdd )
Add(item);
_SuppressCollectionChanged = false;
OnCollectionChangedMultiItem(new NotifyCollectionChangedEventArgs(NotifyCollectionChangedAction.Add, new List<T>(toAdd)));
}
public void Remove(IEnumerable<T> toRemove)
{
if( this == toRemove )
throw new Exception("Invalid operation. This would result in iterating over a collection as it is being modified.");
_SuppressCollectionChanged = true;
foreach( T item in toRemove )
Remove(item);
_SuppressCollectionChanged = false;
OnCollectionChangedMultiItem(new NotifyCollectionChangedEventArgs(NotifyCollectionChangedAction.Remove, new List<T>(toRemove)));
}
#endregion
}
I tackled this one in a slightly different manner as I wanted to register to one event and handle all additions and removals in the event handler. I started off overriding the collection changed event and redirecting reset actions to removal actions with a list of items. This all went wrong as I was using the observable collection as an items source for a collection view and got "Range actions not supported".
I finally created a new event called CollectionChangedRange which acts in the manner I expected the inbuilt version to act.
I can't imagine why this limitation would be allowed and hope that this post at least stops others from going down the dead end that I did.
/// <summary>
/// An observable collection with support for addrange and clear
/// </summary>
/// <typeparam name="T"></typeparam>
[Serializable]
[TypeConverter(typeof(ExpandableObjectConverter))]
public class ObservableCollectionRange<T> : ObservableCollection<T>
{
private bool _addingRange;
[field: NonSerialized]
public event NotifyCollectionChangedEventHandler CollectionChangedRange;
protected virtual void OnCollectionChangedRange(NotifyCollectionChangedEventArgs e)
{
if ((CollectionChangedRange == null) || _addingRange) return;
using (BlockReentrancy())
{
CollectionChangedRange(this, e);
}
}
public void AddRange(IEnumerable<T> collection)
{
CheckReentrancy();
var newItems = new List<T>();
if ((collection == null) || (Items == null)) return;
using (var enumerator = collection.GetEnumerator())
{
while (enumerator.MoveNext())
{
_addingRange = true;
Add(enumerator.Current);
_addingRange = false;
newItems.Add(enumerator.Current);
}
}
OnCollectionChangedRange(new NotifyCollectionChangedEventArgs(NotifyCollectionChangedAction.Add, newItems));
}
protected override void ClearItems()
{
CheckReentrancy();
var oldItems = new List<T>(this);
base.ClearItems();
OnCollectionChangedRange(new NotifyCollectionChangedEventArgs(NotifyCollectionChangedAction.Remove, oldItems));
}
protected override void InsertItem(int index, T item)
{
CheckReentrancy();
base.InsertItem(index, item);
OnCollectionChangedRange(new NotifyCollectionChangedEventArgs(NotifyCollectionChangedAction.Add, item, index));
}
protected override void MoveItem(int oldIndex, int newIndex)
{
CheckReentrancy();
var item = base[oldIndex];
base.MoveItem(oldIndex, newIndex);
OnCollectionChangedRange(new NotifyCollectionChangedEventArgs(NotifyCollectionChangedAction.Move, item, newIndex, oldIndex));
}
protected override void RemoveItem(int index)
{
CheckReentrancy();
var item = base[index];
base.RemoveItem(index);
OnCollectionChangedRange(new NotifyCollectionChangedEventArgs(NotifyCollectionChangedAction.Remove, item, index));
}
protected override void SetItem(int index, T item)
{
CheckReentrancy();
var oldItem = base[index];
base.SetItem(index, item);
OnCollectionChanged(new NotifyCollectionChangedEventArgs(NotifyCollectionChangedAction.Replace, oldItem, item, index));
}
}
/// <summary>
/// A read only observable collection with support for addrange and clear
/// </summary>
/// <typeparam name="T"></typeparam>
[Serializable]
[TypeConverter(typeof(ExpandableObjectConverter))]
public class ReadOnlyObservableCollectionRange<T> : ReadOnlyObservableCollection<T>
{
[field: NonSerialized]
public event NotifyCollectionChangedEventHandler CollectionChangedRange;
public ReadOnlyObservableCollectionRange(ObservableCollectionRange<T> list) : base(list)
{
list.CollectionChangedRange += HandleCollectionChangedRange;
}
private void HandleCollectionChangedRange(object sender, NotifyCollectionChangedEventArgs e)
{
OnCollectionChangedRange(e);
}
protected virtual void OnCollectionChangedRange(NotifyCollectionChangedEventArgs args)
{
if (CollectionChangedRange != null)
{
CollectionChangedRange(this, args);
}
}
}
This is how ObservableCollection works, you can work around this by keeping your own list outside of the ObservableCollection (adding to the list when action is Add, remove when action is Remove etc.) then you can get all the removed items (or added items) when action is Reset by comparing your list with the ObservableCollection.
Another option is to create your own class that implements IList and INotifyCollectionChanged, then you can attach and detach events from within that class (or set OldItems on Clear if you like) - it's really not difficult, but it is a lot of typing.
For the scenario of attaching and detaching event handlers to the elements of the ObservableCollection there is also a "client-side" solution. In the event handling code you can check if the sender is in the ObservableCollection using the Contains method. Pro: you can work with any existing ObservableCollection. Cons: the Contains method runs with O(n) where n is the number of elements in the ObservableCollection. So this is a solution for small ObservableCollections.
Another "client-side" solution is to use an event handler in the middle. Just register all events to the event handler in the middle. This event handler in turn notifies the real event handler trough a callback or an event. If a Reset action occurs remove the callback or event create a new event handler in the middle and forget about the old one. This approach also works for big ObservableCollections. I used this for the PropertyChanged event (see code below).
/// <summary>
/// Helper class that allows to "detach" all current Eventhandlers by setting
/// DelegateHandler to null.
/// </summary>
public class PropertyChangedDelegator
{
/// <summary>
/// Callback to the real event handling code.
/// </summary>
public PropertyChangedEventHandler DelegateHandler;
/// <summary>
/// Eventhandler that is registered by the elements.
/// </summary>
/// <param name="sender">the element that has been changed.</param>
/// <param name="e">the event arguments</param>
public void PropertyChangedHandler(Object sender, PropertyChangedEventArgs e)
{
if (DelegateHandler != null)
{
DelegateHandler(sender, e);
}
else
{
INotifyPropertyChanged s = sender as INotifyPropertyChanged;
if (s != null)
s.PropertyChanged -= PropertyChangedHandler;
}
}
}
Looking at the NotifyCollectionChangedEventArgs, it appears that OldItems only contains items changed as a result of Replace, Remove, or Move action. It doesn't indicate that it will contain anything on Clear. I suspect that Clear fires the event, but does not registered the removed items and does not invoke the Remove code at all.
Well, I decided to get dirty with it myself.
Microsoft put a LOT of work into always making sure the NotifyCollectionChangedEventArgs doesn't have any data when calling a reset. I'm assuming this was a performance/memory decision. If you are resetting a collection with 100,000 elements, I'm assuming they didn't want to duplicate all those elements.
But seeing as my collections never have more then 100 elements, I don't see a problem with it.
Anyway I created an inherited class with the following method:
protected override void ClearItems()
{
CheckReentrancy();
List<TItem> oldItems = new List<TItem>(Items);
Items.Clear();
OnPropertyChanged(new PropertyChangedEventArgs("Count"));
OnPropertyChanged(new PropertyChangedEventArgs("Item[]"));
NotifyCollectionChangedEventArgs e =
new NotifyCollectionChangedEventArgs
(
NotifyCollectionChangedAction.Reset
);
FieldInfo field =
e.GetType().GetField
(
"_oldItems",
BindingFlags.Instance | BindingFlags.NonPublic
);
field.SetValue(e, oldItems);
OnCollectionChanged(e);
}
The ObservableCollection as well as the INotifyCollectionChanged interface are clearly written with a specific use in mind: UI building and its specific performance characteristics.
When you want notifications of collection changes then you are generally only interested in Add and Remove events.
I use the following interface:
using System;
using System.Collections.Generic;
/// <summary>
/// Notifies listeners of the following situations:
/// <list type="bullet">
/// <item>Elements have been added.</item>
/// <item>Elements are about to be removed.</item>
/// </list>
/// </summary>
/// <typeparam name="T">The type of elements in the collection.</typeparam>
interface INotifyCollection<T>
{
/// <summary>
/// Occurs when elements have been added.
/// </summary>
event EventHandler<NotifyCollectionEventArgs<T>> Added;
/// <summary>
/// Occurs when elements are about to be removed.
/// </summary>
event EventHandler<NotifyCollectionEventArgs<T>> Removing;
}
/// <summary>
/// Provides data for the NotifyCollection event.
/// </summary>
/// <typeparam name="T">The type of elements in the collection.</typeparam>
public class NotifyCollectionEventArgs<T> : EventArgs
{
/// <summary>
/// Gets or sets the elements.
/// </summary>
/// <value>The elements.</value>
public IEnumerable<T> Items
{
get;
set;
}
}
I've also written my own overload of Collection where:
ClearItems raises Removing
InsertItem raises Added
RemoveItem raises Removing
SetItem raises Removing and Added
Of course, AddRange can be added as well.
I was just going through some of the charting code in the Silverlight and WPF toolkits and noticed that they also solved this problem (in a kind of similar way) ... and I thought I would go ahead and post their solution.
Basically, they also created a derived ObservableCollection and overrode ClearItems, calling Remove on each item being cleared.
Here is the code:
/// <summary>
/// An observable collection that cannot be reset. When clear is called
/// items are removed individually, giving listeners the chance to detect
/// each remove event and perform operations such as unhooking event
/// handlers.
/// </summary>
/// <typeparam name="T">The type of item in the collection.</typeparam>
public class NoResetObservableCollection<T> : ObservableCollection<T>
{
public NoResetObservableCollection()
{
}
/// <summary>
/// Clears all items in the collection by removing them individually.
/// </summary>
protected override void ClearItems()
{
IList<T> items = new List<T>(this);
foreach (T item in items)
{
Remove(item);
}
}
}
This is a hot subject ... because in my opinion, Microsoft did not do its job properly ... yet again. Don't misunderstand me, I like Microsoft, but they are not perfect!
I read most of the previous comments. I agree with all those who think that Microsoft did not programmed Clear() properly.
In my opinion, at least, it needs an argument to make it possible to detach objects from an event ... but I also understand the impact of it. Then, I thought up this proposed solution.
I hope it will make everybody happy, or at least, most everyone ...
Eric
using System;
using System.Collections.Generic;
using System.Collections.ObjectModel;
using System.Collections.Specialized;
using System.Reflection;
namespace WpfUtil.Collections
{
public static class ObservableCollectionExtension
{
public static void RemoveAllOneByOne<T>(this ObservableCollection<T> obsColl)
{
foreach (T item in obsColl)
{
while (obsColl.Count > 0)
{
obsColl.RemoveAt(0);
}
}
}
public static void RemoveAll<T>(this ObservableCollection<T> obsColl)
{
if (obsColl.Count > 0)
{
List<T> removedItems = new List<T>(obsColl);
obsColl.Clear();
NotifyCollectionChangedEventArgs e =
new NotifyCollectionChangedEventArgs
(
NotifyCollectionChangedAction.Remove,
removedItems
);
var eventInfo =
obsColl.GetType().GetField
(
"CollectionChanged",
BindingFlags.Instance | BindingFlags.NonPublic
);
if (eventInfo != null)
{
var eventMember = eventInfo.GetValue(obsColl);
// note: if eventMember is null
// nobody registered to the event, you can't call it.
if (eventMember != null)
eventMember.GetType().GetMethod("Invoke").
Invoke(eventMember, new object[] { obsColl, e });
}
}
}
}
}
To keep it simple why don't you override the ClearItem method and do whatever you want there ie Detach the items from the event.
public class PeopleAttributeList : ObservableCollection<PeopleAttributeDto>, {
{
protected override void ClearItems()
{
Do what ever you want
base.ClearItems();
}
rest of the code omitted
}
Simple, clean, and contain within the collection code.
I had the same issue, and this was my solution. It seems to work. Does anyone see any potential problems with this approach?
// overriden so that we can call GetInvocationList
public override event NotifyCollectionChangedEventHandler CollectionChanged;
protected override void OnCollectionChanged(NotifyCollectionChangedEventArgs e)
{
NotifyCollectionChangedEventHandler collectionChanged = CollectionChanged;
if (collectionChanged != null)
{
lock (collectionChanged)
{
foreach (NotifyCollectionChangedEventHandler handler in collectionChanged.GetInvocationList())
{
try
{
handler(this, e);
}
catch (NotSupportedException ex)
{
// this will occur if this collection is used as an ItemsControl.ItemsSource
if (ex.Message == "Range actions are not supported.")
{
handler(this, new NotifyCollectionChangedEventArgs(NotifyCollectionChangedAction.Reset));
}
else
{
throw ex;
}
}
}
}
}
}
Here are some other useful methods in my class:
public void SetItems(IEnumerable<T> newItems)
{
Items.Clear();
foreach (T newItem in newItems)
{
Items.Add(newItem);
}
NotifyCollectionChanged(new NotifyCollectionChangedEventArgs(NotifyCollectionChangedAction.Reset));
}
public void AddRange(IEnumerable<T> newItems)
{
int index = Count;
foreach (T item in newItems)
{
Items.Add(item);
}
NotifyCollectionChangedEventArgs e = new NotifyCollectionChangedEventArgs(NotifyCollectionChangedAction.Add, new List<T>(newItems), index);
NotifyCollectionChanged(e);
}
public void RemoveRange(int startingIndex, int count)
{
IList<T> oldItems = new List<T>();
for (int i = 0; i < count; i++)
{
oldItems.Add(Items[startingIndex]);
Items.RemoveAt(startingIndex);
}
NotifyCollectionChangedEventArgs e = new NotifyCollectionChangedEventArgs(NotifyCollectionChangedAction.Remove, new List<T>(oldItems), startingIndex);
NotifyCollectionChanged(e);
}
// this needs to be overridden to avoid raising a NotifyCollectionChangedEvent with NotifyCollectionChangedAction.Reset, which our other lists don't support
new public void Clear()
{
RemoveRange(0, Count);
}
public void RemoveWhere(Func<T, bool> criterion)
{
List<T> removedItems = null;
int startingIndex = default(int);
int contiguousCount = default(int);
for (int i = 0; i < Count; i++)
{
T item = Items[i];
if (criterion(item))
{
if (removedItems == null)
{
removedItems = new List<T>();
startingIndex = i;
contiguousCount = 0;
}
Items.RemoveAt(i);
removedItems.Add(item);
contiguousCount++;
}
else if (removedItems != null)
{
NotifyCollectionChanged(new NotifyCollectionChangedEventArgs(NotifyCollectionChangedAction.Remove, removedItems, startingIndex));
removedItems = null;
i = startingIndex;
}
}
if (removedItems != null)
{
NotifyCollectionChanged(new NotifyCollectionChangedEventArgs(NotifyCollectionChangedAction.Remove, removedItems, startingIndex));
}
}
private void NotifyCollectionChanged(NotifyCollectionChangedEventArgs e)
{
OnPropertyChanged(new PropertyChangedEventArgs("Count"));
OnPropertyChanged(new PropertyChangedEventArgs("Item[]"));
OnCollectionChanged(e);
}
I found another "simple" solution deriving from ObservableCollection, but it is not very elegant because it uses Reflection... If you like it here is my solution:
public class ObservableCollectionClearable<T> : ObservableCollection<T>
{
private T[] ClearingItems = null;
protected override void OnCollectionChanged(System.Collections.Specialized.NotifyCollectionChangedEventArgs e)
{
switch (e.Action)
{
case System.Collections.Specialized.NotifyCollectionChangedAction.Reset:
if (this.ClearingItems != null)
{
ReplaceOldItems(e, this.ClearingItems);
this.ClearingItems = null;
}
break;
}
base.OnCollectionChanged(e);
}
protected override void ClearItems()
{
this.ClearingItems = this.ToArray();
base.ClearItems();
}
private static void ReplaceOldItems(System.Collections.Specialized.NotifyCollectionChangedEventArgs e, T[] olditems)
{
Type t = e.GetType();
System.Reflection.FieldInfo foldItems = t.GetField("_oldItems", System.Reflection.BindingFlags.NonPublic | System.Reflection.BindingFlags.Instance);
if (foldItems != null)
{
foldItems.SetValue(e, olditems);
}
}
}
Here I save the current elements in an array field in the ClearItems method, then I intercept the call of OnCollectionChanged and overwrite the e._oldItems private field (through Reflections) before launching base.OnCollectionChanged
You can override ClearItems method and raise event with Remove action and OldItems .
public class ObservableCollection<T> : System.Collections.ObjectModel.ObservableCollection<T>
{
protected override void ClearItems()
{
CheckReentrancy();
var items = Items.ToList();
base.ClearItems();
OnPropertyChanged(new PropertyChangedEventArgs("Count"));
OnPropertyChanged(new PropertyChangedEventArgs("Item[]"));
OnCollectionChanged(new NotifyCollectionChangedEventArgs(NotifyCollectionChangedAction.Remove, items, -1));
}
}
Part of System.Collections.ObjectModel.ObservableCollection<T> realization:
public class ObservableCollection<T> : Collection<T>, INotifyCollectionChanged, INotifyPropertyChanged
{
protected override void ClearItems()
{
CheckReentrancy();
base.ClearItems();
OnPropertyChanged(CountString);
OnPropertyChanged(IndexerName);
OnCollectionReset();
}
private void OnPropertyChanged(string propertyName)
{
OnPropertyChanged(new PropertyChangedEventArgs(propertyName));
}
private void OnCollectionReset()
{
OnCollectionChanged(new NotifyCollectionChangedEventArgs(NotifyCollectionChangedAction.Reset));
}
private const string CountString = "Count";
private const string IndexerName = "Item[]";
}
http://msdn.microsoft.com/en-us/library/system.collections.specialized.notifycollectionchangedaction(VS.95).aspx
Please read this documentation with your eyes open and your brain turned on.
Microsoft did everything right. You must re-scan your collection when it throws a Reset notification for you. You get a Reset notification because throwing Add/Remove for each item (being removed from and added back to collection) is too expensive.
Orion Edwards is completely right (respect, man). Please think wider when reading the documentation.
If your ObservableCollection is not getting clear, then you may try this below code. it may help you:
private TestEntities context; // This is your context
context.Refresh(System.Data.Objects.RefreshMode.StoreWins, context.UserTables); // to refresh the object context

Resources