How to detect if ASP.NET control properties contain DataBinding expressions? - c#

I have a custom control which inherits from System.Web.UI.Control and some of its properties can be declaratively set using databinding expressions. e.g.
<foo:Foo runat="server" MyFoo="<%# this.GetFoo() %>" />
Now, when I do that I need to call .DataBind() on the control (or one of its parents) to evaluate these expressions.
What I would like to be able to do is detect if any properties were set this way and just automatically have the custom control call this.DataBind() after OnPreRender or there about.
So the question: how do I detect if databinding expressions are waiting to be executed?
I'm convinced that in some ControlBuilder or DataBindContext class lives the information needed to determine this. I've hunted around with Reflector and cannot seem to find it.
I should add, that I don't want to pay the overhead of executing DataBind() if no direct properties have been assigned this way. This is why I'd like to detect before hand. This class is extremely light but I'd like the ability to declaratively set properties without needing any code behind.

Doing some deeper looking into ControlBuilder, I noticed that the compiled factory for each control instance will attach a DataBinding event handler when there are data binding expressions present. I've found that checking for this seems to be a very reliable method for determining if data binding needs to occur. Here is the basis of my solution to the problem:
using System;
using System.Reflection;
using System.Web.UI;
public class AutoDataBindControl : Control
{
private static readonly object EventDataBinding;
private bool needsDataBinding = false;
static AutoDataBindControl()
{
try
{
FieldInfo field = typeof(Control).GetField(
"EventDataBinding",
BindingFlags.NonPublic|BindingFlags.Static);
if (field != null)
{
AutoDataBindControl.EventDataBinding = field.GetValue(null);
}
}
catch { }
if (AutoDataBindControl.EventDataBinding == null)
{
// effectively disables the auto-binding feature
AutoDataBindControl.EventDataBinding = new object();
}
}
protected override void DataBind(bool raiseOnDataBinding)
{
base.DataBind(raiseOnDataBinding);
// flag that databinding has taken place
this.needsDataBinding = false;
}
protected override void OnInit(EventArgs e)
{
base.OnInit(e);
// check for the presence of DataBinding event handler
if (this.HasEvents())
{
EventHandler handler = this.Events[AutoDataBindControl.EventDataBinding] as EventHandler;
if (handler != null)
{
// flag that databinding is needed
this.needsDataBinding = true;
this.Page.PreRenderComplete += new EventHandler(this.OnPreRenderComplete);
}
}
}
void OnPreRenderComplete(object sender, EventArgs e)
{
// DataBind only if needed
if (this.needsDataBinding)
{
this.DataBind();
}
}
}
This solution disables itself if no DataBinding event handler is attached or if the control is manually data bound (directly or via a parent).
Note that most of this code is just jumping through hoops to be able to test for the existence of the event. The only reflection needed is a one-time lookup to get the object used as the key for EventDataBinding.

There is an internal ArrayList called SubBuilders on the ControlBuilder class. For each databinding expression TemplateParser enocunters, ProcessCodeBlock() adds a CodeBlockBuilder object with a BlockType property CodeBlockType.DataBinding to SubBuilders.
So if you can get a handle to the ControlBuilder you want, you should be able to reflectively iterate over SubBuilders and look for objects of type CodeBlockBuilder where BlockType == CodeBlockType.DataBinding.
Note of course this is all kinds of nasty and I'm really suspicious this is the best way to solve your core problem. If you take two steps back and look at the original problem, maybe post that on Stackoverflow instead - there's plenty of super-smart people who can help come up with a good solution.

Related

TwoWay Collection Binding Sync/Lock

What is the best way to synchronize 2 sets of data via Binding?
Target = Custom Setters - raises custom events whenever something changed
Source = ObservableCollection - raises events whenever collection changed
Now my question is, when I receive an update from one collection (e.g. Source.CollectionChanged event) I need to call the custom TargetSetters, and ignore the events called which originated from my update.
And also the other way, when the Target custom events get fired, i need to update the source, but ignore the CollectionChanged event.
At the moment, I am keeping a reference to my handlers, and removing that before updating any of the collections. e.g.
private void ObservableCollection_OnCollectionChanged(object sender, NotifyCollectionChangedEventArgs notifyCollectionChangedEventArgs)
{
CustomObject.SelectionChanged -= CustomObject_SelectionChanged;
// Do change logic and update Custom Object....
CustomObject.SelectionChanged += CustomObject_SelectionChanged;
}
void CustomObject_SelectionChanged(object sender, SelectionChangedEventArgs e)
{
ObservableCollection.CollectionChanged -= ObservableCollection_OnCollectionChanged;
// Do change logic and update ObservableCollection...
ObservableCollection.CollectionChanged += ObservableCollection_OnCollectionChanged;
}
I have seen that you can use an if statement to check if the updates are from source, and if they are ignore them. e.g.
private void ObservableCollection_OnCollectionChanged2(object sender, NotifyCollectionChangedEventArgs notifyCollectionChangedEventArgs)
{
if (BindingTargetUpdating) return;
BindingSourceUpdating = true;
// Do change logic and update Custom Object....
BindingSourceUpdating = false;
}
void CustomObject_SelectionChanged2(object sender, SelectionChangedEventArgs e)
{
if (BindingSourceUpdating) return;
BindingTargetUpdating = true;
// Do change logic and update ObservableCollection...
BindingTargetUpdating = false;
}
After Google + SO Search came back with nothing, I wanted to see how other people are doing this, and is there something really simple I am missing here that solves this problem? (I know that the examples are not thread-safe)
If not, what is the preferred way? Removing and attaching handlers, or setting a boolean flag? What is more performant (yes i know this is highly unlikely to cause a bottleneck but out of curiosity)
Reason I am asking is because, currently I am implementing Attached Behaviours and for each behaviour, I am creating 2 sets of Dictionaries which hold the references to the handlers for each object as state has to be passed around.
I can't seem to find the the source code for the binding mechanism of the .NET Binding classes, to see how MS implemented it. If anyone has a link to those it would be greatly appreciated.
The mechanism you're using - having a boolean which tracks when updates occur, and blocking around it, is the most common approach.
Personally, I prefer to wrap that logic into a small utility that implements IDisposable. This makes it easier to guarantee that you'll always clean up after yourself.
A utility you can use for this would look something like:
class Guard : IDisposable
{
readonly Func<bool> getter;
readonly Action<bool> setter;
readonly bool acquired = false;
public Guard(Func<bool> getter, Action<bool> setter)
{
this.getter = getter;
this.setter = setter;
if (this.getter() == false)
{
this.setter(true);
this.acquired = true;
}
}
public bool Acquired { get { return this.acquired; } }
void IDisposable.Dispose()
{
if (acquired)
{
this.setter(false);
}
}
}
You can then write:
private void ObservableCollection_OnCollectionChanged2(object sender, NotifyCollectionChangedEventArgs notifyCollectionChangedEventArgs)
{
using(var guard = new Guard(() => BindingTargetUpdating, v => BindingTargetUpdating = value))
{
if (guard.Acquired)
{
// Do change logic and update Custom Object....
}
}
}
This isn't necessarily any shorter - its probably longer to write, but does provide guarantees that you'll release your blocks if exceptions occur. You can always subclass Guard to shrink the usage down if you'll be using it frequently.

Use reflection to get actual value of the property notified by INotifyPropertyChanged?

I am working on a project that will use INotifyPropertyChanged to announce property changes to subscriber classes.
void item_PropertyChanged(object sender, PropertyChangedEventArgs e)
{
if (e.PropertyName == "Quantity")
....
It appears to me that when the subscribing class receives the notification, the only available value it can get is the name of the property. Is there a way to get a reference of the actual object that has the property change? Then I can get the new value of this property from the reference. Maybe using reflection?
Would anyone mind writing a code snippet to help me out? Greatly appreciated.
Actual object is sender (at least, it should be):
void item_PropertyChanged(object sender, PropertyChangedEventArgs e)
{
var propertyValue = sender.GetType().GetProperty(e.PropertyName).GetValue(sender);
}
If you care about performance, then cache sender.GetType().GetProperty(e.PropertyName) results.
Note: this interface is primarily a data-binding API, and data-binding is not limited to simple models like reflection. As such, I would suggest you use the TypeDescriptor API. This will allow you to correctly detect changes for both simple and complex models:
var prop = TypeDescriptor.GetProperties(sender)[e.PropertyName];
if(prop != null) {
object val = prop.GetValue(sender);
//...
}
(with a using System.ComponentModel; directive)

Temporarily stop form events from either being raised or being handled?

I have a ton on controls on a form, and there is a specific time when I want to stop all of my events from being handled for the time being. Usually I just do something like this if I don't want certain events handled:
private bool myOpRunning = false;
private void OpFunction()
{
myOpRunning = true;
// do stuff
myOpRunning = false;
}
private void someHandler(object sender, EventArgs e)
{
if (myOpRunning) return;
// otherwise, do things
}
But I have A LOT of handlers I need to update. Just curious if .NET has a quicker way than having to update each handler method.
You will have to create your own mechanism to do this. It's not too bad though. Consider adding another layer of abstraction. For example, a simple class called FilteredEventHandler that checks the state of myOpRunning and either calls the real event handler, or suppresses the event. The class would look something like this:
public sealed class FilteredEventHandler
{
private readonly Func<bool> supressEvent;
private readonly EventHandler realEvent;
public FilteredEventHandler(Func<bool> supressEvent, EventHandler eventToRaise)
{
this.supressEvent = supressEvent;
this.realEvent = eventToRaise;
}
//Checks the "supress" flag and either call the real event handler, or skip it
public void FakeEventHandler(object sender, EventArgs e)
{
if (!this.supressEvent())
{
this.realEvent(sender, e);
}
}
}
Then when you hook up the event, do this:
this.Control.WhateverEvent += new FilteredEventHandler(() => myOpRunning, RealEventHandler).FakeEventHandler;
When WhateverEvent gets raised, it will call the FilteredEventHandler.FakeEventHandler method. That method will check the flag and either call, or not call the real event handler. This is pretty much logically the same as what you're already doing, but the code that checks the myOpRunning flag is in only one place instead of sprinkled all over your code.
Edit to answer question in the comments:
Now, this example is a bit incomplete. It's a little difficult to unsubscribe from the event completely because you lose the reference to the FilteredEventHandler that's hooked up. For example, you can't do:
this.Control.WhateverEvent += new FilteredEventHandler(() => myOpRunning, RealEventHandler).FakeEventHandler;
//Some other stuff. . .
this.Control.WhateverEvent -= new FilteredEventHandler(() => myOpRunning, RealEventHandler).FakeEventHandler; //Not gonna work!
because you're hooking up one delegate and unhooking a completely different one! Granted, both delegates are the FakeEventHandler method, but that's an instance method and they belong to two completely different FilteredEventHandler objects.
Somehow, you need to get a reference to the first FilteredEventHandler that you constructed in order to unhook. Something like this would work, but it involves keeping track of a bunch of FilteredEventHandler objects which is probably no better than the original problem you're trying to solve:
FilteredEventHandler filter1 = new FilteredEventHandler(() => myOpRunning, RealEventHandler);
this.Control.WhateverEvent += filter1.FakeEventHandler;
//Code that does other stuff. . .
this.Control.WhateverEvent -= filter1.FakeEventHandler;
What I would do, in this case, is to have the FilteredEventHandler.FakeEventHandler method pass its 'this' reference to the RealEventHandler. This involves changing the signature of the RealEventHandler to either take another parameter:
public void RealEventHandler(object sender, EventArgs e, FilteredEventHandler filter);
or changing it to take an EventArgs subclass that you create that holds a reference to the FilteredEventHandler. This is the better way to do it
public void RealEventHandler(object sender, FilteredEventArgs e);
//Also change the signature of the FilteredEventHandler constructor:
public FilteredEventHandler(Func<bool> supressEvent, EventHandler<FilteredEventArgs> eventToRaise)
{
//. . .
}
//Finally, change the FakeEventHandler method to call the real event and pass a reference to itself
this.realEvent(sender, new FilteredEventArgs(e, this)); //Pass the original event args + a reference to this specific FilteredEventHandler
Now the RealEventHandler that gets called can unsubscribe itself because it has a reference to the correct FilteredEventHandler object that got passed in to its parameters.
My final advice, though is to not do any of this! Neolisk nailed it in the comments. Doing something complicated like this is a sign that there's a problem with the design. It will be difficult for anybody who needs to maintain this code in the future (even you, suprisingly!) to figure out the non-standard plumbing involved.
Usually when you're subscribing to events, you do it once and forget it - especially in a GUI program.
You can do it with reflection ...
public static void UnregisterAllEvents(object objectWithEvents)
{
Type theType = objectWithEvents.GetType();
//Even though the events are public, the FieldInfo associated with them is private
foreach (System.Reflection.FieldInfo field in theType.GetFields(System.Reflection.BindingFlags.NonPublic | System.Reflection.BindingFlags.Instance))
{
//eventInfo will be null if this is a normal field and not an event.
System.Reflection.EventInfo eventInfo = theType.GetEvent(field.Name);
if (eventInfo != null)
{
MulticastDelegate multicastDelegate = field.GetValue(objectWithEvents) as MulticastDelegate;
if (multicastDelegate != null)
{
foreach (Delegate _delegate in multicastDelegate.GetInvocationList())
{
eventInfo.RemoveEventHandler(objectWithEvents, _delegate);
}
}
}
}
}
You could just disable the container where all these controls are put in. For example, if you put them in a GroupBox or Panel simply use: groupbox.Enabled = false; or panel.Enabled = false;. You could also disable the form From1.Enabled = false; and show a wait cursor. You can still copy and paste these controls in a container other than the form.

ASP.NET dynamic data: Table is not shown anymore after inserting data if FormView.Controls is accessed on Page.Initialized event

Lately, I received a bug report for Ninject.Web that it is not working properly together with ASP.NET dynamic data. The problem is that on postback (e.g. when Inserting, Deleting, Editing a record) the table is not shown anymore.
Some debuging showed that the problem is caused by a IHttpModule that recursively iterates through all controls of a page after it is initialized. As soon as this module accesses the Controls property get accessor of FormView or GridView the problem occurs. If this type of controls is skiped everything is fine. The following code shows the module:
public class NinjectHttpModule : DisposableObject, IHttpModule
{
private HttpApplication httpApplication;
public void Init(HttpApplication context)
{
this.httpApplication = context;
this.httpApplication.PreRequestHandlerExecute += this.OnPreRequestHandlerExecute;
}
private static void InjectUserControls(Control parent)
{
if (parent == null)
{
return;
}
foreach (Control control in parent.Controls)
{
if (control is UserControl)
{
// KernelContainer.Inject(control); This is irrelevant for the question.
}
InjectUserControls(control);
}
}
private void OnPreRequestHandlerExecute(object sender, EventArgs e)
{
var page = this.httpApplication.Context.CurrentHandler as Page;
if (page == null)
{
return;
}
KernelContainer.Inject(page);
page.InitComplete += (src, args) => InjectUserControls(page);
}
}
If this code is changed so that the iteration through the child controls of DataBoundControls is delayed to the DataBound event everything is fine. Shown by the next code snippet:
private static void InjectUserControls(Control parent, bool skipDataBoundControls)
{
if (parent == null)
{
return;
}
if (skipDataBoundControls)
{
var dataBoundControl = parent as DataBoundControl;
if (dataBoundControl != null)
{
dataBoundControl.DataBound += InjectDataBoundControl;
return;
}
}
foreach (Control control in parent.Controls)
{
if (control is UserControl)
{
KernelContainer.Inject(control);
}
InjectUserControls(control, skipDataBoundControls);
}
}
private static void InjectDataBoundControl(object sender, EventArgs e)
{
var dataBoundControl = sender as DataBoundControl;
if (dataBoundControl != null)
{
dataBoundControl.DataBound -= InjectDataBoundControl;
InjectUserControls(dataBoundControl, false);
}
}
Because I'm completely unfamiliar with System.Web.DynamicData I'd like to know some things to get a better feeling about how to fix this bug:
Why does this problem occur? I mean it's only a simple read access to the Controls property.
What side effects can the change above have?
Is it still early enough to inject the controls after the data bound event?
Do you think this is a valid bug fix for this problem?
Certainly puzzling behavior, as can sometimes happen in WebForms with the many phases of execution.
Even though it's just a simple read access to the Controls property, this property can actually do a lot of work to return the child controls. In particular, it can't return the child controls unless they have been created, and that creation normally does not occur until later in the page life cycle. So by accessing it in InitComplete, the children end up getting created prematurely, before some important Dynamic Data hookups have happened, causing some controls to be missing. Yes, I realize that the end result behavior seems to make little sense, which is why some people favor the straightforwardness of MVC :)
As an alternate possible workaround, could you try moving your injection from InitComplete to PreLoad? e.g.
page.PreLoad += (src, args) => InjectUserControls(page);
I'm pretty sure that'll address the problem, though I'm less sure whether this will cause issues with your KernelContainer.Inject logic. Give it a try, since it's simpler than your workaround.
If that doesn't work, I think your workaround is ok, as it delays the enumeration until the children are created. As for 'Is it still early enough to inject the controls after the data bound event', I think that depends in exactly what KernelContainer.Inject does, and what expectations it has on the state of the control.

How can I prevent an event from causing its own event from firing in C#?

I have a treeview with checkboxes and I have the following handler for the "AfterCheck" event:
private void trvAvailableFiles_AfterCheck(object sender, TreeViewEventArgs e)
{
if (!_isCheckingInProgress)
{
trvAvailableFiles.BeginUpdate();
var nodePath = e.Node.Tag.ToString();
bool isChecked = e.Node.Checked;
e.Node.Nodes.Clear();
try
{
_fileTreeLogic.GetChildNodes(e.Node, true);
e.Node.ExpandAll();
_isCheckingInProgress = true;
SetChildrenCheckState(e.Node, isChecked);
_isCheckingInProgress = false;
}
finally
{
trvAvailableFiles.EndUpdate();
}
}
}
If you look closely you'll see that I'm checking if "_isCheckingInProgress". If it is not, then I proceed and expand all the nodes and call the SetChildrenCheckState() method. The problem I have encountered is that SetChildrenCheckState() will subsequently cause each child node to all fire the AfterCheck event for its own node.
My question is, is there a more clean way to allow the first AfterCheck event to fire but not the subsequent ones? It seems kind of hackish that I have to have an instance bool variable to check and set.
Use: if(e.Action != TreeViewAction.Unknown) instead of if (!_isCheckingInProgress). See TreeViewAction .
When the user uses the keyboard or mouse to check the checkboxes, e.Action will be TreeViewAction.ByKeyboard or TreeViewAction.ByMouse.
The MSDN provides this as example code for the TreeView.AfterCheck Event.
Edit 1: Obviously, if you're setting the checkbox yourself in code, move the code in the event handler to a new function and have the code that sets the checkbox call it directly. The point of this solution is to let you use event handlers for user input without having those events get triggered when you set the checkboxes yourself via code.
Edit 2: See Spencer's answer for an explanation of my comment in Edit 1
One recommendation you'll see occasionally around SO is to not put a lot of code into event handlers themselves. There are a number of reasons for this. First, in your case it would be easier to understand a call like:
private void trvAvailableFiles_AfterCheck(object sender, TreeViewEventArgs e)
{
if (!_isCheckingInProgress)
{
_isCheckingInProgress = true;
try { GetAvailableFiles(); } catch {}
_isCheckingInProgress = false;
}
}
And to place the rest of your code in GetAvailableFiles(). This creates a separation between event code and action code which most people would agree is a worthwhile distinction to make.
Second, which may or may not be applicable in your case is that multiple events can cause the same action. Such as mnuFileQuit_Click and btnClose_Click as an obvious example. If both make calls to CloseApplication() it removes a lot of redundant code.
Personally, I use a function that removes and then adds the event.
private void trvAvailableFiles_AfterCheck(object sender, TreeViewEventArgs e)
{
EnableEvents(false);
trvAvailableFiles.BeginUpdate();
var nodePath = e.Node.Tag.ToString();
bool isChecked = e.Node.Checked;
e.Node.Nodes.Clear();
try
{
_fileTreeLogic.GetChildNodes(e.Node, true);
e.Node.ExpandAll();
SetChildrenCheckState(e.Node, isChecked);
}
finally
{
trvAvailableFiles.EndUpdate();
}
EnableEvents(true);
}
private void EnableEvents(bool bEnable)
{
if(bEnable)
cbWhatever.OnChecked += EventHandler;
else
cbWhatever.OnChecked -= EventHandler;
}
No, there's no cleaner way to do what you've shown. I'm not really sure why you feel that variables are a "hack" approach. Setting a flag is a common technique used when writing UI code.
The real hack would be some obscure way to prevent raising the event the first time, but not subsequent times. Future maintenance programmers are guaranteed to understand how setting a flag works; they're not guaranteed to appreciate the "elegance" of your alternative approach.

Categories