Develop Reference

Struggling with WPF internal methods

UPDATE: So, I have a solution to my immediate problem. I haven't succeeded in making "my own" TreeView class. But, the reason I wanted to do that was because controls based on ButtonBase don't function in a Popup from a TreeView, and with the help of #MarkFeldman, I have found a solution that comes at it from a different angle.
The problem is that the MouseDown and MouseUp events bubble, and that bubbling crosses the logical tree boundary between the Popup and its owner. So, when you click on something hosted inside the Popup, the TreeViewItem and TreeView that ultimately own the Popup get to hear about it. This then triggers code inside the TreeView that checks, "Do I have focus?", and if not, helpfully sets focus back to itself -- but being a separate logical tree, the Popup has its own focus context, and so this effectively steals focus from the Button control while it is in the middle of processing a click. The Button responds to this by ignoring the click.
This erroneous handling in the TreeView only happens when MouseDown and MouseUp events reach it. What if there were a way to prevent it from seeing those events in the first place? Well, if you intercept the PreviewMouseDown and PreviewMouseUp events and mark them Handled, then the framework doesn't generate MouseDown and MouseUp events to begin with.
Looking at the Reference Source, it looks like ButtonBase's click handling is tied up in a couple of protected methods:
https://referencesource.microsoft.com/#PresentationFramework/src/Framework/System/Windows/Controls/Primitives/ButtonBase.cs,414
https://referencesource.microsoft.com/#PresentationFramework/src/Framework/System/Windows/Controls/Primitives/ButtonBase.cs,478
This means you can call them from your own subclasses! So, instead of making "my own" TreeView where all controls behave properly, instead I can make "my own" CheckBox that works properly in a Popup from a TreeView. Since all of the actual click handling is directly accessible, and the events it normally responds to use the same EventArgs type as the Preview events, and on top of it the default handling takes care of marking the events as Handled, the entire implementation boils down to this:
public class CheckBoxThatWorks : CheckBox
{
protected override void OnPreviewMouseLeftButtonDown(MouseButtonEventArgs e) => base.OnMouseLeftButtonDown(e);
protected override void OnPreviewMouseLeftButtonUp(MouseButtonEventArgs e) => base.OnMouseLeftButtonUp(e);
}
Nice!
ORIGINAL QUESTION:
I need to make a clone of the TreeView WPF class -- a copy of the control that runs out of my own code. (There is a bug in the control and Microsoft doesn't seem to deem it high-enough priority to fix, and it makes it completely impossible to host buttons (including check boxes and radio buttons) within pop-ups shown from TreeViewItems. link)
I am running into serious difficulties with the amount of internal shenanigans Microsoft has undertaken in the implementation of base WPF controls. A couple of issues I've bumped into:
Controls have a property HandlesScrolling that allows them to take over basic scrolling handling from ScrollViewer. But, it's marked internal, making it seemingly impossible for me to have a control of my own that does its own handling of scrolling from keyboard input. I was going to try having my TreeView handle keyboard scrolling in OnPreviewKeyDown instead of OnKeyDown, so that it can prevent KeyDown events from being raised for the keys it intercepts. I haven't gotten far enough to know what caveats there might be about this.
The Visual States system allows you to declare what styles should be applied when different states are entered, but actually entering states seems to be tied up in the virtual method ChangeVisualState on the Control type. All controls that want to switch between visual states can override this method and inspect their state to determine which Visual State should be shown. Oh wait. They can't because the method is internal! Apparently only Microsoft gets to create controls that set their own visual states??
Are there any strategies I can use to work around these limitations, or am I just completely out of luck?

How can I create dynamic controls in response to a button click?

I want to fill an updatepanel with new dynamic controls in response to a button click.
However, I also want to be able to access the values of these dynamic controls in response to an event in one of the controls.
Specifically, I want the button to bring up two dropdownmenus. One of the menus (or both if need be) is in another update panel. I want the first menu in the update panel to change its data in response to a value getting selected in the other menu.
I think my problem is that when I cause a postback with one dropdownmenu I lose the other dropdownmenu because I created it in the button_click handler.
I know I should create dynamic controls in the Page_Init method (or so ive heard) but I only want the controls to show up if the button is clicked. There are other buttons on the page which need to create a different set of dynamic controls.
Thanks.

There are a lot of ways you can handle this, and which approach to take really depends on your project's requirements and your available resources.
The smoothest way to do it that would generally provide the best user experience would be to use a Javascript technique to hide and show controls as the page required them. JQuery is the library I would recommend for this. On the most basic level, you simply wire the control's activation (such as a button_click event) and hide or show a div containing the dynamic content as necessary, like so:
$("#control").show();
// and
$("#control").hide();
Alternatively, you can do this in C# by using the Visible property on many of the normal web controls for ASP.NET. The usual code-behind approach would look something like this:
private void btnControl_Click(object sender, EventArgs e)
{
var dynamicControl1 = FindControl("dynamicControl1");
dynamicControl.Visible = false; // or true, as the case may be
}
This particular approach is mostly attached to code-behinds, though, which I would encourage you to avoid if possible. They are practically impossible to test and will make projects a pain to work in. You can use a similar approach in the MVC3 framework, of course, it will just be a little different how you send and receive the control you are setting to not be visible. The other benefit this has that is kind of nice is that if something is set to not be visible, it tends not to even be displayed in the HTML generated by the templating engine (YMMV depending on the engine, but I know this is true in Razor). So someone viewing the source of your webpage won't be able to see inactive controls, which may or may not be something that appeals to you.
EDIT: I see the problem is less to do with how to display these things, and more with how to create and read them back given on-the-fly input.
I'm sure there's a way to do this with Javascript (which would more than probably be the cleanest and best way to do this), but I'm not good enough with JS to know the answer to that one. The way you would handle this in ASP.NET is make the div you're going to add controls to server-side (by using runat='server', then add what you need there. Again, the trivial code-behind approach would be something like:
private void btnControl_Click(object sender, EventArgs e)
{
foreach(var checkBoxChecked in chkBoxes.Where(x => x.Checked))
{
div.Controls.Add(new WebControl()) // or whatever the heck else it is you need.
}
}
This presumes that you have an IEnumerable<CheckBox> to iterate over, of course. You may also want an IList<WebControl> to keep track of all the junk you're adding. You will also need to make sure the CSS is applied properly to the div for the controls you're adding. And again, code-behinds are pretty awful and I use the example only because it'd be easy to spin up in a project to test for yourself.

.Net TableLayoutPanel – Clearing Controls is Very Slow

This is really simple.
I have a TableLayoutPanel that is populated with controls (just Labels, Buttons, and some Panels with buttons) based on a database query. When the data needs to be refreshed, I use TableLayoutPanel.Controls.Clear(). Unfortunately, this is a very slow operation. I would expect it to be faster than the code populating the table, but it is at least 3 or 4 times slower.
I definitively proved that the slowness is when executing Controls.Clear() by executing this as the single thing done to the TableLayoutPanel after a message box is displayed (then the procedure returns). The controls visibly disappear from the bottom up. When the recordset is used to repopulate the TableLayoutPanel, the speed of the controls appearing from top to bottom is almost faster than I can see.
I'm already doing TableLayoutPanel.SuspendLayout() and ResumeLayout().
Using this.DoubleBuffered = true on the form doesn't appear to do anything.
I could just Dispose the entire control and recreate it through code, but this is a big pain and makes having a nice form designer GUI pointless. I would have to dig into every property I've set on the control and create a line of code for it (though I guess I could get this out of the designer code itself, it still feels wrong).
Any ideas on how to do the job faster? I'm even open to using other methods besides a TableLayoutPanel... I just need the freedom to put multiple buttons per cell or barring that to be able to span columns in the table header.
Can C# at least freeze the whole form while it redraws and then paint all at once?

I've run into issues with slowness using TableLayoutPanels as well. Rather than setting the DoubleBuffered property on the form, the best solution I have found is to create a new class that inherits from TableLayoutPanel, and in that class' constructor, enable double-buffering:
public class DoubleBufferedTableLayoutPanel : TableLayoutPanel
{
public DoubleBufferedTableLayoutPanel()
{
DoubleBuffered = true;
}
}
Then, use the DoubleBufferedTableLayoutPanel wherever you would normally use a TableLayoutPanel.

This seems to work for my uses:
tableLayoutPanel.Visible = false;
tableLayoutPanel.Clear();
/* Add components to it */
tableLayoutPanel.Visible = true;

There is no need to subclass TableLayoutPanel as in Chris Ryan's answer. I had the same problem and solved it by setting the property through reflection:
typeof(TableLayoutPanel)
.GetProperty("DoubleBuffered",
System.Reflection.BindingFlags.NonPublic | System.Reflection.BindingFlags.Instance)
.SetValue(myTableLayoutPanel, true, null);

If i'm going to built up some dynamic gui i'm always going to do so in code. But at a starting point i just start with the designer on a dummy form and style each control the way i (or better the customer) like(s). Afterwards i take a look into the Designer.cs file and copy the necessary property settings out of it into some factory function like
private TextBox CreateTextBox(string name, /* maybe other parameters */)
{
var textBox = new TextBox();
textBox.Name = name;
//Other settings from given parameters...
//Further settings which are all the same for these kind of control
textBox.KeyDown += (sender, e) => {};
return textBox;
}
So i make sure that every control feels and looks the same on my GUI. This will be done on each level within my surface (starting with the small controls like TextBox and goes up to the containers like GroupBox or TableLayoutPanel.
In some cases this leads to a point where a factory function calls several other factory functions. If this is becoming true it's time to think about encapsulating these controls into a single UserControl, but as always it depends if this is needed or not.
From my side i can only encourage you to move your code out of the designer into a self-written function. At the beginning it is (as always) more work, but afterwards it is easier to make even bigger changes to the layout.

2 or more Controls only 1 is considered Active

I have 2 controls (MyCtrl) next to each other called ctrlLeft and ctrlRight.
Whenever one receives interaction it is considered active (by default the left one). I override OnDraw and customize the look of the active one a bit.
Currently I have a property Active and an event that I subscribe to from all MyCtrl in there I store a reference to the active one like this:
if (sender is MyCtrl)
{
ctrlActive = (sender as MyCtrl);
ctrlLeft.Active = !(ctrlRight.Active = (ctrlActive == ctrlRight));
}
Either way I need to have ctrlActive as I use it for other things but what I am wondering is if this is the best way make them aware of each other?
Another option I thought of was to store references to every possible MyCtrl and then loop through em all and activate / deactivate the one that match sender just in case I in the future add a ctrlMiddle.
Are my thoughts wrong, is there better options to do this. For example, how does radiobuttons accomplish their similar functionality?
Edit: Thanks for all suggestions.
In my particular case I don't want/need a container as even if I have 3 MyCtrl only one can still be active and I don't want them to be "linked" 2 and 2 so I went with a public static MyCtrl { get; set; } that each control can check itself against and I can update it where I need to, which works in my case and rids me of the need to loop through a collection when using multiple MyCtrl.

Your methods are sound. If you need multiple controls with only one being active consider a container (visual or otherwise) where the children supply an "activate" event TO the container and the children also subscribe to a "control activated" event FROM the container....if the child is not the control supplied by the containers "control activated" event..then paint it as not active otherwise paint as active.

You have to manage the activation by yourself, so I think that your method is good.

I nearly had the same idea as Rusty. But i would make it a little more general.
Why not building a container control that contains two panels (like SplitContainer). The container control has a property Active of type enum ActivePanel { First, Second } and it can be switched from outside by a setter or automatically through a subscription of the container to the Focus event (or something similar).
With such an approach you also don't need a ctrlMiddle cause you can nest your container multiple times.
I think there are still some problems to solve by this idea, but it should give you a good direction.

How bad is the following snippet?

My question is simple: how bad is the following snippet of code? How would you do it?
CancelEventHandler _windowClosing;
private CancelEventHandler WindowClosing
{
set
{
clearEventHandlerList();
this.Closing += value;
_windowClosing = value;
/*
* if calling the method with null parameters,
* it will set up itself as the primary control on the Window
*/
_windowClosing(null,null);
}
get
{
return _windowClosing;
}
}
private readonly CancelEventHandler[] CONTROLS = null;
private int current = 0;
public InitializerForm()
{
InitializeComponent();
/*
* these are the handlers for the different controls,
* in the order of appereance to the user
*/
STATES = new CancelEventHandler[] { handler1, handler2, etc. };
WindowClosing = CONTROLS[0];
}
private void clearEventHandlerList()
{
foreach (CancelEventHandler c in CONTROLS)
{
this.Closing -= c;
}
}
private void handler1(object obj, CancelEventArgs e)
{
if (obj == null)
{
//hide every other control, but this one, also set up if necessary
}
else
{
//do something
WindowClosing = CONTROLS[++current]; // set the next control to show
e.Cancel = true;
}
}
The point would be that the code wouldn't close a form, but instead show another component on it, and the set the way to handle that (this is mobile platform, so clicking OK button on the top generates a closing event). This is because showing several forms (4 or 5) one after another to the user makes the app blink, and also very annoying, while replacing just components is much smoother. This model works, but seems very nasty, and I would like a cleaner way to handle this.
Update:
I updated the code sample so that variable names are somewhat speaky. Still, I'm convinced this is awful, (a) but not sure how much, and more importantly, (b) how to do it better.
Update 2:
So, it seems that the code is still a bit mysterious.
Now here's what the problem is:
I show the user a form, which instructs him what to do in several languages. He proceeds by clicking OK on the window. Next, I ask for his language, and then a few questions (where his/her GPS is, etc.) like this. After he could answer the questions (this shouldn't take more than a few seconds each), I show him a splash screen (I load stuff in a separate thread meanwhile), which has a picture. Showing these forms one after another makes the whole application start slow, and filled with UI lags.
Here's what I do to work around the lags: I put the content of the windows into panels, and put those panels one on another, and hide every one of them but the one that should be visible to the user. (current variable) Each of the windows does different things, so I need to change handler of the window closing event in addition. In this code the part which enables the panel is in the same function (handler1, handler2, etc.) with the part which handles the window closing event. If the arguments are null, it does the former, if it isn't (that means it was triggered by the user) it does the latter.
I need an extensible solution to this so that I can insert and remove dialogs anytime I want (the order and the pointers to the functions are stored in the CONTROLS field, and this seems to be very convenient, if you actually understand it. Although it is never easy to change the entire content of a form, there ought to be a simpler way to do this, as well a nicer one, that is what I'm looking for.
I hope this time I could explain how the model works.

I think it might be theoretically possible to make that code more delightfully diverting, perilously puckish, jovially jeopardous, cheerily chancy and unwarily whimsical but it would require some serious thought.

somehow your code makes me want to cry, i´m sorry. i read it twice and all i know about it is that it "doesStuff" with "STATES".
if you really want some help on this one you will have to work on it yourself first...

Use, XML! It's human-readable!
More seriously-
It seems like you're trying to create some sort of configuration wizard, so I'd start by researching that. Regarding your particular solution, I generally advocate very strongly against the "layered panel" approach. I do so because I maintain apps written by people who found this approach, or the related "hidden tabs on a tab control" approach, to be a good idea. It's not, and maintainers will curse your name for years to come.
That being said, what alternatives are there? Well, one alternative is what you've already dismissed because of its "flicker". I'd say that, in general, the flicker isn't that big of a deal for a quick and dirty application. It might be a good idea to make sure that your new window is called up before closing the old one, for example. (I'm assuming this is possible, I haven't developed on a mobile device.)
Another possibility might be a less-evil version of your layered panels. Instead of throwing a half-dozen panels into one form, create a separate user control for each wizard page and then add/remove the user controls to a containing form. This can avoid your flicker and will prove to be much easier to maintain because each page is in a different control. This might also ease any subsequent "Back" button functionality and make your data structures more naturally defined because those user controls will be associated with a specific logical bit of data. It's still not ideal, but it's probably good enough for a one-off solution.
A third technique, if you foresee extensive wizard modification as the product matures, might be to generalize the creation of your user controls by defining them in a more logical/declarative manner (e.g. via XML). If you dynamically generate sensible controls based on XML, then modifying the panels might be as easy as diving into your XML and doing something like:
<Questions>
<Question type="Text"> <!-- generate a textbox for the answer field -->
Favorite Color:
</Question>
<Question type="Number" range="0-255"> <!-- Maybe this is a spinner -->
The answer to life, the universe, and everything:
</Question>
</Questions>
That's just off the top of my head, and completely overkill for any one-off application, but it's a possibility.

Now, let me caveat this by saying this might work, but it may not be the answer to your real problem - that of a slow and unresponsive UI when you have a lot of forms. The real answer may be to just go ahead and do all separate forms, but have each form load its child forms in a background thread while the user is staring at the first form.
But assuming you're still set on this, I'd start off by making a separate class just to handle the Panel stacking/hierarchy. Call it PanelManager. You would instantiate the PanelManager and associate it with the main form, then add Panels to it (perhaps keyed to a String) and set the order. In the main form, have the closing handler call PanelManager.CloseCurrentPanel() and if there are no more Panels to show then it's time to close the main form.
Time for pseudo-code! Here's a quick idea for the class, i'll leave it to you to implement it:
public class PanelManager {
// constructor
public PanelManager (Form ownerForm);
// short-cut properties
public Panel this[int idx]
{ get; set; }
public int Index
{ get; set; }
// main functionality
public int AddPanel (Panel p);
public void SetPanelOrder (Panel p, int idx);
public void RemovePanel (Panel p);
public void RemovePanelAt (int idx);
// shows the first Panel
public void Show ();
// shows Panel[idx]
public void Show (int idx);
// adds the panel to the top of the stack and displays it
// returns the index of the panel
public int AddPanelAndShow (Panel p);
// hides the current panel, displays the one underneath it
// returns false if there are no more panels
public bool HideCurrentPanel ();
}
in the constructor for the main form, instantiate it by new PanelManager (this), then in the closing event handler, call panelManager.HideCurrentPanel () and then figure out whether or not you need to close it after that.