Gdiplus::Bitmap *WindowsSystem::GetIconPixelData(HICON hIcon)
{
	ICONINFO iconInfo;
	GetIconInfo(hIcon, &iconInfo);

	BITMAP iconBmp;
	GetObject(iconInfo.hbmColor, sizeof(BITMAP),&iconBmp);

	Gdiplus::Bitmap* bitmap = new Gdiplus::Bitmap(iconBmp.bmWidth,iconBmp.bmHeight,PixelFormat32bppARGB);

	bool hasAlpha = false;
	{
		// We have to read the raw pixels of the bitmap to get proper transparency information
		// (not sure why, all we're doing is copying one bitmap into another)

		Gdiplus::Bitmap colorBitmap(iconInfo.hbmColor, NULL);
		Gdiplus::BitmapData bmpData;
		Gdiplus::Rect bmBounds(0, 0, colorBitmap.GetWidth(), colorBitmap.GetHeight());

		colorBitmap.LockBits(&bmBounds, Gdiplus::ImageLockModeRead, colorBitmap.GetPixelFormat(), &bmpData);
		for (int y = 0; y < colorBitmap.GetHeight(); y++)
		{
			byte *pixelBytes = (byte*)bmpData.Scan0 + y*bmpData.Stride;

			for (int x = 0; x SetPixel(x, y, Gdiplus::Color(*pixel));
				hasAlpha = hasAlpha || (pixelBytes[3] > 0 && pixelBytes[3] < 255);
			}
		}
		colorBitmap.UnlockBits(&bmpData);
	}

	if (!hasAlpha)
	{
		// If there's no alpha transparency information, we need to use the mask
		// to turn back on visible pixels

		Gdiplus::Bitmap maskBitmap(iconInfo.hbmMask,NULL);
		Gdiplus::Color cMask, cBitmap;
		for (int y = 0; y < maskBitmap.GetHeight(); y++)
		{
			for (int x = 0; x GetPixel(x, y, &cBitmap);
					cBitmap.SetValue(cBitmap.GetValue() | 0xFF000000); // turn alpha to opaque (i.e. 0xFF)
					bitmap->SetPixel(x, y, cBitmap);
				}
			}
		}

	}

	return bitmap ;

}

The first problem is that, if the background is not set to anything, there will be no events transmitted. A simple solution is just to set the background to transparent or white (or whatever color you want).

The second problem, which took me forever to figure out, stems from the fact that the Canvas width was 1 pixel (or was it 0?) once I put it into my StackPanel. However, this does not keep elements in the Canvas from appearing outside this 1-pixel wide region, so it appears as if the width of the Canvas is greater. So the important thing to realize is that the Canvas will only respond to events that are within the area covered by its width/height.

It seems that other people have the same problem as well.

The solution is to add another node in the hierarchy, so you have:

(your custom class) –> (a private container class) –> Child

Then, you could just modify the container class’s transforms, and it would affect the presentation of the Child. I coded up a trivial example which scales everything 2x:

using System;
using System.Windows;
using System.Windows.Controls;
using System.Windows.Media;
using System.Windows.Shapes;
namespace Model
{
    class ZoomAndPan : Decorator
    {
        Decorator _container;

        public ZoomAndPan()
        {
            _container = new Decorator();
            AddVisualChild(_container);
            AddLogicalChild(_container);

            _container.RenderTransform = new RotateTransform(45);
        }

        public override UIElement Child
        {
            get { return _container.Child; }
            set { _container.Child = value; }
        }

        protected override int VisualChildrenCount
        {
            get { return _container != null ? 1 : 0; }
        }

        protected override Visual GetVisualChild(int index)
        {
            if (index > 0 || _container == null)
                throw new ArgumentOutOfRangeException("index");

            return _container;
        }

        protected override Size MeasureOverride(Size sizeAvailable)
        {
            _container.Measure(sizeAvailable);
            return _container.DesiredSize;
        }

        protected override Size ArrangeOverride(Size sizeArrange)
        {
            _container.Arrange(new Rect(new Point(0, 0), sizeArrange));
            return sizeArrange;
        }

        [STAThread]
        static void Main(string[] args)
        {
            Rectangle r = new Rectangle();
            r.Fill = Brushes.Bisque;
            r.Width = 200;
            r.Height = 200;

            Canvas canv = new Canvas();
            canv.Children.Add(r);

            ZoomAndPan zp = new ZoomAndPan();
            zp.Child = canv;

            Window w = new Window();
            w.Content = zp;

            new Application().Run(w);
        }
    }
}

For the purposes of a project I was working on, I needed to detect when the user was pressing a key or moving the mouse. Because my project is all done in C#, I found this piece of code from Code Project very helpful. It’s just a drop-in class that notifies you whenever you receive a keystroke or mouse activity. All was great except for one thing: when you click on the minimize, maximize, or close buttons of your application, the whole application starts stuttering. Other applications running at the time are not affected, but this little problem plagues every implementation of global mouse hooks that I could find, whether written in managed or unmanaged code. Instead, I’m going to use a more roundabout method. I’ll have a very small Win32 app print basic information about events to the console, and then read this information in from my C# application. I found this great little C++ global hooking app after a lot of looking.

Addendum (10/6/2007):

It turns out  a much simpler way is to set up the hook in a new thread. You simply create a new thread and call SetWindowsHookEx from that thread. However, if that is all you do you will not be receiving any events. You need to create a message loop in the thread that reads messages coming to it and processes them. Fortunately, this is very easy. Using WPF, all you need to do is call Dispatcher.Run() after you initialize the hook. Using Windows Forms, I think all you need to do is call Application.Run().

The Mono Project has C# bindings for cairo, but unfortunately the latest version still didn’t support everything I needed. I was fortunate, however, to come across cleaned-up cairo-sharp library at NDesk. I took the latest snapshot, and because I don’t have make installed on my computer, I simply created a Microsoft Visual C# Express class library project, put all the files inside of it, and built the Mono.Cairo.dll assembly.

There were several links that were especially pertinent to what I was trying to do. Cairo has a guide to using cairo and OpenGL, and the Mono Project had some information about using the Mono.Cairo bindings. Unfortunately, their links to more samples did not seem to work.

In the end, I was able to get a very simple example working. The steps are after the jump:

1. If you haven’t already, install the Tao Framework.
2. Make a copy of the cleaned-up cairo-sharp (i.e. Mono.Cairo) library from NDesk. You will also need the cairo dlls; in my case, they were already installed as part of Mono.
3. We will be modifying a Tao Framework / NeHe sample project. Go to your Tao Framework installation, and go to the source\examples\NeHe directory.
4. Create a new C# Windows Application in Visual Studio and replace all the code in the file Program.cs with the text from the file Lesson06.cs in the Tao Framework NeHe samples directory.
5. Add a reference to the Tao Framework OpenGL Binding and Windows Platform API Binding. Add a reference to the Mono.Cairo assembly as well. Add the line “using Cairo;” at the top of the source file. A few classes will now be ambiguous when you recompile; fix them to use the standard class library (not Cairo).
6. Change the name of the method Run to Main. In case you’re curious, you can compile but the program will crash because it cannot find a texture that it tries to preload.
7. At the beginning of the LoadGLTexture method, add the following code (taken from the Mono website). This code generates an in-memory bitmap for Cairo and draws something interesting on it.
   

   Cairo.ImageSurface i = new Cairo.ImageSurface(Format.Argb32, 256, 256);
Cairo.Context gr = new Cairo.Context(i);

   // Shape
gr.MoveTo(new PointD(100, 200));
gr.CurveTo(new PointD(100, 100), new PointD(100, 100),
new PointD(200, 100));
gr.CurveTo(new PointD(200, 200), new PointD(200, 200),
new PointD(100, 200));
gr.ClosePath();

   // Save the state to restore it later. That will NOT save the path
gr.Save();
Cairo.Gradient pat = new Cairo.LinearGradient(100, 200, 200, 100);
pat.AddColorStop(0, new Cairo.Color(0, 0, 0, 1));
pat.AddColorStop(1, new Cairo.Color(1, 0, 0, 1));
gr.SetSource(pat);

   // Fill the path with pattern
gr.FillPreserve();

   // We "undo" the pattern setting here
gr.Restore();

   // Color for the stroke
gr.SetSource(new Cairo.Color(0, 0, 0));

   gr.LineWidth = 3;
gr.Stroke();

8. Remove the instantiation for textureImage and replace all references to textureImage with i (i is the Cairo bitmap). Completely remove the follow two lines:
   

   textureImage[0] = LoadBMP("NeHe.Lesson06.NeHe.bmp"); // Load The Bitmap
BitmapData bitmapData = textureImage[0].LockBits(rectangle, ImageLockMode.ReadOnly, PixelFormat.Format24bppRgb);

9. The most important step. Previously the program was copying an imported file into a texture. Now we want to copy out Cairo bitmap instead. We replace this line:
   Gl.glTexImage2D(Gl.GL_TEXTURE_2D, 0, Gl.GL_RGB8, textureImage[0].Width, textureImage[0].Height, 0, Gl.GL_BGR, Gl.GL_UNSIGNED_BYTE, bitmapData.Scan0);
   with this one:
   Gl.glTexImage2D(Gl.GL_TEXTURE_2D, 0, Gl.GL_RGB8, i.Width, i.Height, 0, Gl.GL_BGRA, Gl.GL_UNSIGNED_BYTE, i.Data);
10. At the end of the method, before we return, insert calls to dispose the Context and ImageSurface objects we created:
    

    ((IDisposable)gr.Target).Dispose();
((IDisposable)gr).Dispose();

11. Voila! You have pretty vector graphics being mapped onto a cube.
    

You can get the Microsoft Visual C# Express project as well.