I'm working in .NET, C# to be specific, creating a Win Forms UserControl, which contains a WebBrowser control. The WebBrowser control hosts a page, which in turn uses a third-party javascript component. The problem I'm having is with invoking a javascript function to initialize the third-party javascript component and block the UI in the Windows Forms application until the component has been initialized, which the component notifies you of through an internal javascript event that it has.
Part of the problem is that the only way to change any configuration parameter of the third-party javascript component is to re-initialize it with the new configuration. So for example, if you want to make it read-only you have to re-initialize it with the read-only parameter.
I've got everything working in terms of being able to call the Document.InvokeScript and then in the web page call the UserControl method using window.external but the problem I'm having is how to block the UserControl code that makes the call to initialize the javascript component so that it waits and doesn't return control to the user until the initialization of the javascript component has been completed.
The reason I need it to work this way is because if I have a "Read-Only" checkbox on the form that changes the the ReadOnly property of the UserControl to control whether the javascript component shows the data as read-only and the user clicks that checkbox really quickly you will either get a javascript error or the checkbox will get out of sync with the actual read-only state of the javascript component. This seems to happen because the control hasn't re-initialized yet after it's configuration has changed and you're already trying to change it again.
I've spent hours and hours trying work out a way to make it work using everything from AutoResetEvent to Application.DoEvents and so on, but don't seem to be able to get it working.
The closest I've found is Invoke a script in WebBrowser, and wait for it to finish running (synchronized) but that uses features introduced in VS2012 (and I'm using VS2010) and I don't think it would work anyway as it's a bit different in that you're not waiting for a javascript event to fire.
Any help would be greatly appreciated.
The problem in the first place is the requirement to "block" the UI thread until some event has been fired. It's usually possible to re-factor the application to use asynchronous event handlers (with or without async/await), to yield execution control back to the message loop and avoid any blocking.
Now let's say, for some reason you cannot re-factor your code. In this case, you'd need a secondary modal message loop. You'd also need to disable the main UI while you're waiting for the event, to avoid nasty re-entrancy scenarios. The waiting itself should to be user-friendly (e.g., use the wait cursor or progress animation) and non-busy (avoid burning CPU cycles on a tight loop with DoEvents).
One way to do this is to use a modal dialog with a user-friendly message, which gets automatically dismissed when the desired JavaScript event/callback has occured. Here's a complete example:
using System;
using System.Runtime.InteropServices;
using System.Windows.Forms;
namespace WbTest
{
[ComVisible(true)]
[ClassInterface(ClassInterfaceType.None)]
[ComDefaultInterface(typeof(IScripting))]
public partial class MainForm : Form, IScripting
{
WebBrowser _webBrowser;
Action _onScriptInitialized;
public MainForm()
{
InitializeComponent();
_webBrowser = new WebBrowser();
_webBrowser.Dock = DockStyle.Fill;
_webBrowser.ObjectForScripting = this;
this.Controls.Add(_webBrowser);
this.Shown += MainForm_Shown;
}
void MainForm_Shown(object sender, EventArgs e)
{
var dialog = new Form
{
Width = 100,
Height = 50,
StartPosition = FormStartPosition.CenterParent,
ShowIcon = false,
ShowInTaskbar = false,
ControlBox = false,
FormBorderStyle = FormBorderStyle.FixedSingle
};
dialog.Controls.Add(new Label { Text = "Please wait..." });
dialog.Load += (_, __) => _webBrowser.DocumentText =
"<script>setTimeout(function() { window.external.OnScriptInitialized}, 2000)</script>";
var canClose = false;
dialog.FormClosing += (_, args) =>
args.Cancel = !canClose;
_onScriptInitialized = () => { canClose = true; dialog.Close(); };
Application.UseWaitCursor = true;
try
{
dialog.ShowDialog();
}
finally
{
Application.UseWaitCursor = false;
}
MessageBox.Show("Initialized!");
}
// IScripting
public void OnScriptInitialized()
{
_onScriptInitialized();
}
}
[ComVisible(true)]
[InterfaceType(ComInterfaceType.InterfaceIsIDispatch)]
public interface IScripting
{
void OnScriptInitialized();
}
}
Which looks like this:
Another option (a less user-friendly one) is to use something like WaitOneAndPump from here. You'd still need to take care about disabling the main UI and showing some kind of waiting feedback to the user.
Updated to address the comment. Is your WebBrowser actually a part of the UI and visible to the user? Should the user be able to interact with it? If so, you cannot use a secondary thread to execute JavaScript. You need to do it on the main thread and keep pumping messages, but WaitOne doesn't pump most of Windows messages (it only pumps a small fraction of them, related to COM). You might be able to use WaitOneAndPump which I mentioned above. You'd still need to disable the UI while waiting, to avoid re-entrancy.
Anyhow, that'd still be a kludge. You really shouldn't be blocking the execution just to keep the linear code flow. If you can't use async/await, you can always implement a simple state machine class and use callbacks to continue from where it was left. That's how it used to be before async/await.
Related
I have a web browser frame docked in a Silverlight application, and sometimes full-window XAML/Silverlight UI elements can pop up over it. The old problem, which I've more-or-less fixed, is that the web frame's contents didn't seem to play nice with the Silverlight content, always wanting to be rendered on top. This isn't really an issue now since I have an event that fires whenever an application-obscuring popup appears, but now there's a new problem.
My code that launches pop-ups looks like this:
public void OnModuleShown()
{
if (ModuleShown != null)
ModuleShown(this, new ModuleShownEventArgs());
}
public void ShowModule (string uri, string headerTitle, string message, string transition = "DefaultTransition")
{
// Security and destination validation
GetInfoFromURI(uri, out contentKey, out dataKey, out securityToken);
OnModuleShown();
ShowLoadingSpinner();
_loadModule(contentKey, dataKey, securityToken, headerTitle, message, transition);
}
My code that handles the event looks like this:
private void Shell_ModuleShown(object sender, ModuleShownEventArgs e)
{
if (browserFrame.Visibility == Visibility.Visible)
{
browserFrame.Visibility = Visibility.Collapsed;
return;
}
}
The new problem is that even though I call the event before I start loading and displaying the new module, and even though all the event is doing is changing a web frame's visibility, the module tends to appear first if the loading time is short. Since the module's appearance is animated, it looks even worse since the web frame seems to be waiting for the module to finish its animation before it vanishes.
Questions
Is there some kind of threading method I can use to address this? I really don't want to use Thread.Sleep but it's the only one I know of that would fix this without large program changes I can't make. Even better would be if there was a way to get this web frame to play along with Z-indexes or something similar.
I am using Visual Studio 2013, and my project's .NET Framework version is 4.0.
I have a WPF project and from the main window i am creating and loading some bunch of user controls, there is some large data i am loading in background and then updating a built-in control throw the dispatcher, that works fine, the problem is that some of the user controls loads a lot of data, for example the very first thing i load in the main area of my main window, what i want is to put a loading label instead, load the main window as fast as possible so the user see this label and run in background the creation of that user control and when is done add it as a child of my main container area on my main window while i remove the loading label, if i follow the same philosophy i run into the same error like when i run a task and then try to update the window without using the dispatcher. i want to be able of create the user control asynchronous then update the main window.
Code:
User Control:
public partial class CustomUserControlGallery : UserControl
{
public CustomUserControlGallery()
{
InitializeComponent();
}
...
}
On the backend class of the main window:
public partial class MainWindow : Window
{
CustomUserControlGallery _customUserControlGallery;
public MainWindow()
{
InitializeComponent();
Task t = new Task({
//Can't use the _customUserControlGallery's Dispatcher because object is uninitialized and this.Dispatcher not working either.
_customUserControlGallery = new CustomUserControlGallery(); //Error Here.
_gridContainer.Dispatcher.Invoke(new Action(() => _gridContainer.Children.Add(_customUserControlGallery)));
_loadingLabel.Visbility = Visibility.Collapse;
});
t.Start();
}
...
}
I don't know how to handle this situation with the thread associated to the user control and the main thread.
Error:
{"The calling thread must be STA, because many UI components require this."}
You're doing this wrong. All controls must be created & operate on the UI Thread. That said, you can use the BackgroundWorker class to load the data.
You typically do this by disabling the control whose data is being loaded in the background or hiding it & displaying a progress indicator in its place. Then, you start your BackgroundWorker. That can communicate how far along it is using the ReportProgress method. Finally, when it's finished running, the RunWorkerCompleted event is fired, and you use that to either enable the control, or to hide the progress indicator & show the control.
Some quick & dirty (untested) code:
Place this in your Initialize() or control constructor:
private BackgroundWorker loadData = new BackgroundWorker();
loadData.DoWork += loadData_DoWork;
loadData.ProgressChanged += loadData_ProgressChanged; // Only do this if you are going to report progress
loadData.WorkerReportsProgress = true;
loadData.WorkerSupportsCancellation = false; // You can set this to true if you provide a Cancel button
loadData.RunWorkerCompleted += loadData_RunWorkerCompleted;
private void DoWork( object sender, DoWorkEventArgs e ) {
BackgroundWorker worker = sender as BackgroundWorker;
bool done = false;
while ( !done ) {
// If you want to check for cancellation, include this if statement
if ( worker.CancellationPending ) {
e.Cancel = true;
return;
}
// Your code to load the data goes here.
// If you wish to display progress updates, compute how far along you are and call ReportProgress here.
}
}
private void loadData_ProgressChanged( object sender, ProgressChangedEventArgs e ) {
// You code to report the progress goes here.
}
private void loadData_RunWorkerCompleted( object sender, RunWorkerCompletedEventArgs e ) {
// Your code to do whatever is necessary to put the UI into the completed state goes here.
}
What you are essentially saying (I think) is that Your app becomes sluggish while your control renders a large amount of data.
This is a problem that needs to be solved via virtualisation. You cannot create a control on a background thread, have it render its data behind the scenes and then pop it into existence. You can create controls on separate dispatchers, but they cannot share the same visual and logical tree, so you will not be able to have one as a child of the other.
Virtualisation is what you need to focus on. Depending on the control you can use a variety of virtualisation settings. Try googleing the subject as there is a lot of information on how to achieve this effectively. Most likely you will want to use things like virtualizing stackpanels and container recycling.
You cannot create UI controls with different Dispatchers and use them with each other. It's just not allowed. What you want to do is on your Task you do the heavy lifting work without UI updates and when it is done you push it back to the Dispatcher to update the UI.
In your case, I wouldn't even use Dispatcher.Invoke. Since you are using Task, it has a TaskScheduler.FromCurrentSynchronizationContext() that you can pass in the constructor.
What is the purpose of instantiating controls in a different thread if you're just going to put it back to the Main dispatcher? It's not expensive to do that.
I'm modifying existing WinForms project. The project has UserControl.
This UserControl has DataSet variable which is set from another part of the program in different thread.
What I want to do is to dynamically add another controls to this control depending on the DataSet.
So, after DataSet is loaded, I'm calling RefreshChildControl function and trying to add my new ChildUserControls to flowLayoutPanel. And that's where the problems begin:). I get the "Cross-thread operation not valid: Control 'ChildUserControl' accessed from a thread other than the thread it was created on" exception. I tried to use if(this.InvokeRequired) and Invoke this method, but it does not help. InvokeRequired on MyUserControl is false.
So, is there any good way of performing such task? Or am I missing something important?
EDIT:
I tried to skip InvokeRequired test and just call this.FindForm().Invoke on this method. I've got "Invoke or BeginInvoke cannot be called on a control until the window handle has been created." exception. And, by the way, when I open another form with this control everything worked fine.
First. The simplest solution is to perform Invoke everytime. Nothing bad will happen.
Second, use SynchronizationContext.
using System.Threading;
public class YourForm
{
SynchronizationContext sync;
public YourForm()
{
sync = SynchronizationContext.Current;
// Any time you need to update controls, call it like this:
sync.Send(UpdateControls);
}
public void UpdateControls()
{
// Access your controls.
}
}
SynchronizationContext will manage all threading issues for you. It checks, whether you call from the same or from the other thread. If from same it will just immediately execute your code. Otherwise it will do Invoke through form's message loop.
If your user control is not immediately visible after you construct it, the handle will not be created on the thread that you think it is created on. It's not the C# object whose thread parent is important, it is the Windows Handle object whose parent is important.
To force a control to be immediately created on the thread that you thought you created it on, then
read out the control.Handle which will force the control to actually be made and assigned a handle.
MyUserControl uc = new MyUserControl(); // the handle is not created here
uc.Visible = false;
IntPtr dummy = uc.Handle; // The control is immediately given a real handle
You can also try to fiddle around with uc.CreateControl, but this won't create the handle if the control is not visible.
Now you can have another thread update your user control even if the user control is not visible.
uc.BeginInvoke((Action)(() => uc.Text = "ha ha"));
If you leave out the dummy = uc.Handle line, you will get an exception that you can't call BeginInvoke on a control that does not have a handle.
http://msdn.microsoft.com/en-us/library/system.windows.forms.control.createcontrol(v=vs.90).aspx
I am working on a text editor that is based on RichEditBox. I have implemented functionality "Go to line" which eventually resolves to
TextPointer.Paragraph.BringIntoView();
Along with this I also set the caret position.
What I found out is that BringIntoView only works when I click on the RichEditBox first (focus it). Otherwise it seems to get ignored. I can see that the caret position has been adjusted by the code around BringIntoView though.
Does anybody know what is the reason/nature of that problem? How can I overcome it?
Found a workaround for this, not sure if it will work in a pure WPF environment, in my case I'm running WPF inside a mainly Windows Forms solution using WPF UserControls where needed.
Instead of invoking BringIntoFocus() immediately, defer it to a later moment by adding it to a queue that gets handled by a timer. For example:
System.Windows.Forms.Timer DeferredActionTimer = new System.Windows.Forms.Timer() { Interval = 200 };
Queue<Action> DeferredActions = new Queue<Action>();
void DeferredActionTimer_Tick(object sender, EventArgs e) {
while(DeferredActions.Count > 0) {
Action act = DeferredActions.Dequeue();
act();
}
}
In your forms constructor, or in the OnLoad event add:
DeferredActionTimer.Tick += new EventHandler(DeferredActionTimer_Tick);
DeferredActionTimer.Enabled = true;
Finally, instead of calling TextPointer.Paragraph.BringIntoView(); directly, call it like this:
DeferredActions.Enqueue(() => TextPointer.Paragraph.BringIntoView());
Note that the Windows Forms timer kicks events off in the main thread (via the message pump loop). If you have to use another timer you need a bit of extra code. I'd recommend you to use System.Timers.Timer rather than the System.Threading.Timer (it's a little more thread-safe). You would also have to wrap the action in a Dispatcher.Invoke structure. In my case, the WinForms timer works like a charm.
Can't you just give the RichTextBox(?) focus first then, using Keyboard.Focus(richTextBox) or richTextBox.Focus()?
Short version: I want to trigger the Form_Load() event without making the form visible. This doesn't work because Show() ignores the current value of the Visible property:
tasksForm.Visible = false;
tasksForm.Show();
Long version: I have a WinForms application with two forms: main and tasks. The main form is always displayed. The user can either click a button to open the tasks form, or click some buttons that just run a task directly without opening the tasks form.
When a user asks to run a task directly, I'd like to just call some public methods on the tasks form without showing it. Unfortunately, the task logic depends on stuff that happens in the Form_Load() event. The only way I can find to trigger Form_Load() is to call Show(). The best I've been able to do is to show the form in the minimized state:
tasksForm.WindowState = FormWindowState.Minimized;
tasksForm.Show();
I suppose the cleanest solution would be to pull the tasks logic out of the tasks form and into a controller class. Then I can use that class from the main form and from the tasks form, and only load the tasks form when I need it visible for the user. However, if it's an easy thing to load the form without displaying it, that would be a smaller change.
Perhaps it should be noted here that you can cause the form's window to be created without showing the form. I think there could be legitimate situations for wanting to do this.
Anyway, good design or not, you can do that like this:
MyForm f = new MyForm();
IntPtr dummy = f.Handle; // forces the form Control to be created
I don't think this will cause Form_Load() to be called, but you will be able to call f.Invoke() at this point (which is what I was trying to do when I stumbled upon this SO question).
It sounds to me like you need to sit down and re-think your approach here. I cannot imagine a single reason your public methods need to be in a form if you are not going to show it. Just make a new class.
I totally agree with Rich B, you need to look at where you are placing your application logic rather than trying to cludge the WinForms mechanisms. All of those operations and data that your Tasks form is exposing should really be in a separate class say some kind of Application Controller or something held by your main form and then used by your tasks form to read and display data when needed but doesn't need a form to be instantiated to exist.
It probably seems a pain to rework it, but you'll be improving the structure of the app and making it more maintainable etc.
From MSDN:
Form.Load
Occurs before a form is displayed for the first time.
Meaning the only thing that would cause the form to load, is when it is displayed.
Form.Show(); and Form.Visible = true; are the exact same thing. Basically, behind the scenes, Show checks for various conditions, then sets Visible to true. So obviously, setting visible to false (which it already is) before showing the form is meaningless.
But let's forget the technicalities. I completely agree with Rich B and Shaun Austin - the logic shouldn't be in that form anyway.
Sometimes this would be useful without it being bad design. Sometimes it could be the start of a migration from native to managed.
If you were migrating a c++ app to .NET for example, you may simply make yourwhole app a child window of the .NET form or panel, and gradually migrate over to the .NET by getting rid of your c++ app menu, status bar, toolbar and mapping teh .NEt ones to your app using platform invoke etc...
Your C++ app may take a while to load, but the .NET form doesn't..in which you may like to hide the .NEt form until your c++ app has initialised itself.
I'd set opacity=0 and visible=false to false after calling show, then when your c++ app loads, then reverse.
If you make the method public, then you could access it directly.... however, there could be some unexpected side effects when you call it. But making it public and calling it directly will not draw the screen or open the form.
Move mandatory initialization code for the form class out of the Load event handler into the constructor. For a Form class, instantiation of an instance (via the constructor), form loading and form visibility are three different things, and don't need to happen at the same time (although they do obviously need to happen in that order).
None of the answers solved the original question, so, add the below, call .Show() to load the form without showing it, then call .ShowForm() to allow it to be visible if you want to after:
private volatile bool _formVisible;
protected override void SetVisibleCore(bool value)
{
base.SetVisibleCore(_formVisible);
}
public void ShowForm()
{
_formVisible = true;
if (InvokeRequired)
{
Invoke((Action) Show);
}
else
{
Show();
}
}