Develop Reference

How to make a window manager?

I tried writing code several different times, but I came to an error with each one.
Basically, I'm trying to make "windows" similar to say Explorer, Paint, MediaPlayer, where you could drag then around, interact with them, minimize and close. Of course, if you clicked on a window, the one below it (they can overlap) shouldn't get affected.
I know how to do this, I have a list of the class I call Window, loop through it, and I only interact with the first window to contain the location of the mouse-click. This way, other windows overlap won't get affected.[1]
Next, I had to make it so that two buttons that are overlapping don't get activated when the user clicks in the "intersection of both buttons." I handled this by using the same method I used above.[2]
But the problem I'm facing now is that, if I hold the left click, but then I decide not to click a button, I drag the mouse away from the button, and release the left click, so that the button-click event won't be activated. But, when I remove the mouse from the boundaries of the button, and say, into another.. the new button get activated. Which it should not.[3]
My set up is like this:
I have a class called Window.
In Window, I have a list of the class called Interface (similar to the Control class in WinForms).
And each Interface has a struct in it that contains 4 bools, if the left/right is currently down, and if they were down in the previous processing. (prevLeft, prevRight, currLeft, currRight)
So, I'm ready to discard that (I have not yet, so I still have the source code), but I need a good structure for making an object-oriented type of application. However, I am not using WinForms. I need help with the structure alone, so no actual code is necessary, description is enough. I need to avoid the 3 problems I mentioned above.

Creating your own Window Manager is not an easy task. I know it because I'm making one too ;)
You can use an existing, though maybe not the best solution, like for example Nuclex.UI, which I personally rejected when I first saw it, but if you're not dead set on making your own WM, I suggest to use that or hybrid WinForms-XNA approach.
But if you're really dead set on implementing a custom Window Manager, you have to understand how any other WM works. Since we're talking about XNA, it means Windows, and that means Windows Explorer, which is a great thing to learn from.
You have to recognize how the simplest things work, and it's really not so hard. The hard part is figuring out what logic is updated when, and how to not spend all the CPU on only UI updates. Let me just give you a few hints on how to solve the problems you mention in your question.
To keep track of all windows, I'm using a Dictionary<string, Window>, where Window is a custom class, and the string is its unique name for rare cases where I have to call windows by name. Think of it as a window GUID or Handle. But you can just make it so that a "Form" can only appear once, and store all references in static variables.
To make WM understand what control you're clicking I use rectangles and check if they contain a Point which is at Cursor coordinates and has {1; 1} pixel size, which is probably about the same way it's done in Windows Explorer. To do that your WM needs to know in which order to update the active windows. Usually you'd want to start from the topmost window and continue towards the end of the list of active windows. For that you can just iterate through the list with a foreach loop.
But that's not all, because every window itself is a Container, which means it contains other controls, some of which may even be Containers themselves, like WinForms Panel class. This means you have to iterate through each of the Windows' Children controls. The update order should make sense too - update from the topmost child to the bottommost, recursively for Container controls, in case they also have Containers in them. This basically means you'd want to implement a recursive GetAllControls() method for your WindowManager class that would iterate through all Containers and return a list of all Controls.
Drawing all those Controls should be done in reverse order of updating them, so you can just GetAllControls().Reverse() and iterate through that in a foreach loop.
Where to draw and what to update depends on all the parent containers the current container has and their combined offset from the top-left corner of the game window. I solve this by storing a ParentContainer reference in all children controls to get the appropriate DrawRectangles and update areas via recursive properties.
When you click somewhere on the screen and a click is registered on a Control, make the WindowManager remember that (bool clickRegistered) and not run any OnClick events on any underlying Controls.
Windows Explorer remembers the control you clicked and will activate its OnRelease event if the cursor is then released in the update area of the very same control. So basically Windows Manager only does something when you release the mouse button. You can make your WindowManager and Controls to handle click events differently, like firing an event right after you press the mouse button, i.e. OnMouseDown. But remember that Microsoft aren't noobs and there's a reason for that behavior in Windows Explorer, and it's because if you accidentally press a mouse button somewhere you didn't intend, you can still fix it by moving the cursor outside the pressed control's update area and not run its action.
At this point you might be thinking "Is it really worth implementing all this?" For me the answer was "maybe", because I was a total noob in both C# and XNA at the time I started, and now I know my game, which was originally supposed to use some Window Manager, is going to benefit from my own WM implementation far more than from ready third-party solutions. And besides, it's a great exercise in logic and programming.
But if you'd like to think of yourself as a game developer, you should think in terms of reaching your goal as quickly as possible, i.e. actually making a game, and not the game engine. So in this case, better make use of existing solutions and start selling your product.

Instead of having the structure with the 4 booleans (similar to xna), how about you make a way to tell where the mouse "is." So in a sense, the mouse is in Window number 5 which is Paint, and the user is holding the mouse down on interface/control number 2 which is a button.
That sounds like it could work.

Read item being dragged by mouse

I'm wondering if their is any way to see what's currently being dragged by the mouse. I don't mean over a winforms as i can handle events and get it that way but has anyone been able to invoke some of the win api to read the object or information about it?
I'm trying 'monitor' (probably not the best choice of words) the cursor and see whats being dragged and then potentially read that object.
C# / C++ idea's all welcome !
Thanks in advance

One way to do this by design is to inject code into all applications, by means of a hook.
Using Hooks (Windows): http://msdn.microsoft.com/en-us/library/windows/desktop/ms644960(v=vs.85).aspx
This will allow you to detect when dragging is occurring, and you can use the standard windows APIs that the application itself can use to find out what is being dragged.
SetWindowsHookEx: http://msdn.microsoft.com/en-us/library/windows/desktop/ms644990(v=vs.85).aspx
A second way is to use Windows UI automation. This will not give you exactly what the application sees, or give you access to the exact data being dragged or dropped, but it may give you enough information for whatever your purposes are.
UI Automation support for drag and drop: http://msdn.microsoft.com/en-us/library/windows/desktop/hh707386(v=vs.85).aspx
Try using UISpy or Inspect.exe to see UI Automation events.
https://stackoverflow.com/questions/1848721/where-do-i-get-ui-spy

C# Suppressing mouse click for multiple mice

I have written an application that currently handles clicks from multiple mouse devices.
I used this project and modified to handle mouse clicks as apposed to keyboards.
This is working fine, however now I need to know if there is a way to suppress a click event if it has been handled by my app. The app is a quiz game so the idea is that the quiz master will have (and still be able to use) 1 mouse, and the other contestants will have their own mouse (as buzzers). So when they buzz in, I don't want the mouse click events to fire in the operating system (or at least this application).
The concept is the familiar override of void WndProc(ref Message message), and so I have tried not calling base.WndProc(ref Message) when I don't want the click events to fire, but this has not worked.
Can anybody point me in the right direction here?
Should I be going down the windows hook route? I have looked at this, but I can't seem to work out how I could hook to each mouse device individually.
Any help would be greatly appreciated.
Edit:
This is a Windows Form UI project, and not WPF. So the MultiPoint SDK from Microsoft won't work.

The solution to this lies within not WndProc, but PreFilterMessage(). By intercepting messages before they even reach the form, you can remove them from the message pump causing them to never reach the control that was clicked. This also works for child controls within the form.
I answered this and posted the full source in the following question:
C# Get Mouse handle (GetRawInputDeviceInfo)

I want to drag an image from a Silverlight UIElement and drop it on the users Desktop or Windows Explorer

I have a Silverlight control containing an image. I want the user to be able to drag the image out of the Silverlight application and drop it anywhere they would be able to drop an image. For example, to the Desktop or to a PowerPoint slide or Word document. Everything that I have read thus far says it cannot be done but I find that hard to believe. I'm very new to Silverlight and RIA development so any help would be much appreciated.
Below is the code sample in my WinForm Form but the drag never starts.
string[] aString = { imagePath };
DataObject data = new DataObject(DataFormats.FileDrop, aString);
data.SetData(DataFormats.StringFormat, imagePath);
DoDragDrop(data, DragDropEffects.Copy);

Well the trouble is that a drag operation in Silverlight doesn't have simple access to anything outside the browser (by design). Depending on the user's settings you even have to get explicit permission for clipboard operations and sandboxed temporary file storage. This really sounds like a task better suited to a WPF application (perhaps with web deployment?) or some other desktop application technology.
However, that being said here are some things you could try/consider:
Silverlight/Javascript/ActiveX combination hosted in Internet Explorer
Silverlight 5 "Out of Brower" & P/Invoke (I heard P/Invoke will be supported when Silverlight 5 comes out)
Silverlight connecting to a web service running on the same computer (crazy, but you didn't ask for "not crazy", you asked for possible)
I am not very familiar with drag and drop in the Win32 API so it would take a lot of research and experimentation before I could confirm that this was even possible (and I can already tell you it isn't practical).
Edit: Based on the extra information you provided about the question I suspect it is possible to do what you are attempting. First, are you using WPF or WinForms? I assume WPF but one of your comments says WinForms. I wasn't very familiar with WPF drag/drop operations, but having looked into it, I think your code is on the right path. I created a WPF application and initiated a drag during a KeyDown event. This meant that the mouse button was not necessarily pressed. If I initiated the DragDrop while the button was down it worked. If I initiated while the mouse button wasn't down, I had to push the mouse button down and the drag operation would start (this was unexpected since I assumed the mouse would have to already be down). If I pressed the mouse down outside the application, then gave the WPF app focus (ALT+Tab), then initiated the DragDrop while the mouse button was still down, it didn't work. I got a reference to the MouseDevice and checked the LeftButton property, and the state was showing as "Released" even though the button was still being held down. It seems the key here is the way drag/drop interacts with internal mouse state. You might have to find a way to set the mouse state (maybe with the UI Automation API?). At this point it should be painfully obvious that this whole thing is a hack (even though it is probably possible to get it to work somehow).

The solution we came up with was as follows. The RIA i.e. Silverlight sends a message to our Desktop application WinForms with the path of the image to drag along with the bounding rectangle in screen coordinates that we want to start the drag from. The Desktop code creates and places a Panel over the area that we want to drag from. This panel is where we use DoDragDrop to initiate the native drag when the user left clicks. Since this panel is placed outside and above the silverlight control, everything works perfect. Sandbox defeated.

I have been working on some Silverlight apps for the past few months and fully investigated your exact requirements only to find it was not possible. I believe you can drag from the file system, from Silverlight control to control, but not to the file system.
Does Silverlight 4 support drag and drop from app to desktop?
http://msdn.microsoft.com/en-us/library/dd772166%28v=vs.95%29.aspx

Is there a way to ask about the status of a drag operation while it is in progress?

I know that when a drag/drop operation is completed, upon receiving a MouseUp or Esc key event, it returns an enum that indicates what happened (Move, Copy, None, etc.) My question is this: is there a way to send back status information to the form/control that initiated the drag event, while it is going on?
The use case is as follows (think Visual Studio-esque layout manager for all of this): I am writing a layout/window managing component that allows regions of the layout to be dragged around. I use a transparent form to paint a semi-transparent overlay that changes based on where the mouse is dragging over, a la the preview overlay that appears when dragging windows around in Visual Studio.
Another motivation is that the serialization process I describe is relatively resource intensive, and I'd prefer not to do it if the dragging is all going to occur within the same process/window. So, if there was a way to lazily serialize only when an actual "drop" in another window happens, that would probably make all the difference in usability.
What I want to do is enable dragging between different windows or even different instances of the application. I've already plumbed out the serialization code and everything, but the issue is that, when I drag a chunk of layout into another window, the first window doesn't have any way of knowing that the mouse is now over another instance of the application, which is more than capable of painting its own overlay. So, the original overlay hangs around like an idiot and my program looks like crap.
Is there any way for me to pass along some kind of callback or is there any message or property I can listen for/poll during a drag operation that will tell me if my mouse pointer is over a region that can accept its data? Please don't make me resort to listening for the CursorChanged event, I've already lost too much self respect using reflection to hack around weird wpf/winforms dragging interop bugs. If anyone could suggest a clean resolution for this problem I would be extremely grateful.
Additionally, if anyone could point me to any favorite sites which describe how to go about doing reeeeally funky things with drag and drop, it would be appreciated, as I've found there is quite a lack of really nitty gritty information available about dragging. Usual things like custom cursors and the like are okay, but I'm probably more interested in Win32 black arts and the like.
UPDATE:
I actually just found out about the GiveFeedback event a second ago, came back to my question, and there it was. Huge facepalm moment. However, since I've got you here, what about my second question: is there any way to lazily load the information only when it encounters a valid target? Could I somehow implement my own IDataObject or do things get marshaled righ when the mouse leaves the form? GiveFeedback provides me only with whether there's a valid target under the cursor, but doesn't let me change what data is being dragged...
ANOTHER UPDATE:
Is there any way to determine the source of a drag operation? That is, when my control receives a DragEnter message, how can I tell if the source of the drag is my own control or a foreign one? I know I can hackishly encode it by messing with the AllowedEffects property, but is there any more direct route?

Check out the GiveFeedback event (there's a nice article here) - that sounds to me to be exactly what you're after.