Feeds:
Posts
Comments

Archive for the ‘Software’ Category

Some days ago I read  Jon Skeet‘s twitter post on NCrunch and thought I would give it a try.

The idea of running unit tests right from the IDE without having to call NUnit or any other third party testing framework seems interesting.

To test NCrunch, I reused my GIS solution that has some unit tests.

  • Opened my solution in VS 201o
  • Enabled NCrunch from VS 2010 main menu
  • NCruch wizard popped up

At this point, I had to go back to the code as I needed to do some modifications before running the wizard. Unfortunately that was not possible as the wizard is a modal dialog so I had to cancel it.

Once my modifications on the code done, I tried to go back and open the wizard again by disabling and then enabling NCrunch.

That did not have any effect and I could not reopen the wizard.

As I could not get the wizard back for my GIS solution, I decided to go for a new project from scratch.

Enabling NCrunch for a new project does not work either, not sure but I was expecting the wizard to open for every new solution created within VS2010.

Anyway, the tools looks very simple and easy to configure… trying without the wizard!

I created a test project that contains a simple class that does some math operations, the example below is a method that adds up two integer numbers and returns the result.

public class MathOperations {
   public static int Add(int i, int j) {
       return i+j;
   }
}
[TestFixture]
class MathOperationsTest {
    [Test]
    public void TestAddition() {
        Assert.AreEqual(2, MathOperations.Add(1, 1));
    }
}

Note that if you want to have your tests automatically executed , you need to “enable automatic test execution” from the main NCrunch menu.

NCrunch shows you the test results straight in the IDE which is awesome! No need to go anywhere else to see your results as they will be there sitting right in front of you along with the production code!

The tool uses the standard colors to highlight test results, all successfully tested lines are marked in green, and that happens in real time!

Up to this point, the tool looks great, so far so good! Time to push it a bit into the limits…

For this purpose, the addition method showed above is rewritten using recursion, which involves a more complex execution path and more resources.

   public static int Add(int i, int j) {
      if (j == 0) return i;
      return Add(i+1, j-1);
   }
   [Test]
   public void TestAddition() {
      Assert.AreEqual(300, MathOperations.Add(100, 200));
   }

The resulting output is a bench of green dots, meaning that the production code is doing what it is supposed to do.

Changing the unit test to the following :

   [Test]
   public void TestAddition() {
      Assert.AreEqual(2, MathOperations.Add(1000000, 500000));
   }

Gives… A crash 😦

In this situation, it is preferable for the tool to show a dialog box warning the user of a memory exception rather than crashing.

To wrap up this post, some points to be highlighted:

The good 

  • NCrunch is an excellent tool that helps you test your code without having to leave your IDE. This will save you a couple of clicks and that is something like 5 seconds of your precious time for every test execution you perform
  • There are many other useful features available not covered in this post, check out  NCrunch web site for more details

The bad

  • The tool being in its beta release, there are still some reliability issues that need to be fixed

Still in the process of trying out this tool, I will post any other findings I do while testing my different sized projects.

Happy testing!

Read Full Post »

WPF vs HTML5

An interesting question a friend of mine asked a couple of days ago on Twitter, would HTML5 beat WPF?

In this short post, I’ll try to answer this question and share my personal thoughts on this very interesting matter. One thing I would like to mention here is that my knowledge in WPF, HTML5 are not strong and wide enough to compare these two domains, that’s why I will be speaking my opinion from a technology point of view not technical.

In my mind, no technology is lasting forever. Our excitement for today’s innovations will fade away tomorrow as the industry involves and new technologies come to market. Just think for a second, how excited were you the day you learned that MS is releasing a new Win32 technology called MFC and how disappointed were you when you got the sad news of MS letting down MFC for other technologies such as WinForms? Every technology comes and makes a lot of success in a defined era and then goes away. This being said, I am pretty sure many WPF principles and paradigms sush as MVVM, data binding, UI/domain layers decoupling can be reused in different technologies. One thing for sure, HTML5 might have a chance to knock down WPF but that will only happen on the web development ring, standalone applications can still be built with this later technology and have a great success on the market.

Read Full Post »

The most important part of my drawing engine is almost done. In this post, I will talk about a coding mistake that took me almost a day to resolve hoping this will save some precious time to others. I spent a couple of hours adding a new renderer to my GIS application capable of drawing custom data layers. Done with implementing all drawing steps and got to display the final result and… Nothing on the screen!

Here is a code sample, part of the new rendering class

public DrawingVisual Render(Geometry geo){
      DrawingVisual drawingVisual = new DrawingVisual();
      DrawingContext drawingContext = drawingVisual.RenderOpen();
      // Draw geometry
      drawingContext.DrawGeometry(Brushes.Red, new Pen(Brushes.Blue, 2), geom);
      return drawingVisual;
}

This bit of code does obviously not work! Every thing seems to be right in place but something hidden somewhere is wrong making the shapes not to appear on the screen.
In fact, there was one line missing! The drawing context was left open and that is the reason why my drawings were not reflected on the screen.
I fixed the issue by adding the following line at the end of the “Render” method.

      drawingContext.Close();

Read Full Post »

Hi,

Now that I could see a visual representation of my project, next is sharpening the architecture and design of the application.

One design pattern that comes  very often when reading about WPF is MVVM (http://en.wikipedia.org/wiki/Model_View_ViewModel). I used this pattern and separate my GUI business from my domain objects, I struggled a bit at the beginning but finally managed to get something out.

Before moving to the steps taken to get an MVVM architecture out of my application, let’s have a quick look at the current implementation. In my previous post I directly used the map object from SharpMap and displayed it in the GUI as an image:

1- Load a shape file using SharpMap API

2- Render the shape file into an image

3- Use the resulting image and display it directly in the main window

Here is the pieces of code used for this purpose

// Event handler for button click
public void OpenShapeFileClicked(object sender, EventArgs e){
      // ... ask the user to provide a shape file using a file open dialog control
      LoadShapeFile(path_provided_by_user);
}
public void LoadShapeFile(string fileName){
      SharpMap.Layers.VectorLayer myLayer = new SharpMap.Layers.VectorLayer("ShapeFiles");
      myLayer.DataSource = new SharpMap.Data.Providers.ShapeFile("path_to_file", true);
      myLayer.Style.Fill = System.Drawing.Brushes.GreenYellow;
      myLayer.Style.Outline = new System.Drawing.Pen(System.Drawing.Brushes.Gray);
      myLayer.Style.Line.Width = 2;
      myLayer.Style.Line = new System.Drawing.Pen(System.Drawing.Brushes.Black);
      myMap.Layers.Add(myLayer);
      myMap.ZoomToExtents();
      DrawImage();
}
private void DrawImage(){
      // Convert System.Drawing.Image to ImageSource to wich an image is bound
      using (System.Drawing.Bitmap bitmap = new System.Drawing.Bitmap(myMap.GetMap())){
             IntPtr hBitmap = bitmap.GetHbitmap();
             // Use System.Windows.Interop.Imaging.CreateBitmapSourceFromHBitmap
             BitmapSource bitmapSource = CreateBitmapSourceFromHBitmap(
                hBitmap,
                IntPtr.Zero,
                Int32Rect.Empty,
                System.Windows.Media.Imaging.BitmapSizeOptions.FromEmptyOptions());
             image1.Source = bitmapSource;
      }
}

This is the usual way most programmers will do and especially those writing WinForms applications. Obviously not  the way to go as the domain code is linked to the GUI which makes it very hard to read and maintain.

What we want here is certainly not coupling the GUI with the business domain layer. In order to achieve separation of concerns, we will make use of the MVVM pattern and data binding.

Before looking on how to bind the map to the main GUI, let’s first have a look at the different artifacts of the application and where do they fit in MVVM:

Model : the map data structure, contains GIS data organized in layers as per the GIS simple feature specifications

View : a very thin layer communicating directly to the WPF main window

ViewModel: A new class that is introduced to separate the view from the model

Let’s see this pattern in action in a real life example: loading shape files! Let’s look at how shape files are loaded after moving the application’s architecture to MVVM.

The main control worth mentioning here is the Canevas that will hold the map. This control is updated upon model change notifications.

In order to be aware of all domain objects data changes, the main window will subscribe to change notifications coming from the map model (domain objects):

// Listen to model changes
mapView.PropertyChanged += new System.ComponentModel.PropertyChangedEventHandler(MapViewModelPropertyChanged);

This bit of code is placed on the main window constructor. This is a simple way of subscribing to all model data changes, albeit not a perfect one ( WPF encourages binding to the model directly on the XAML)

The ‘Load shape file’ button on the main window is bound to a command on the view class that will take care of loading all needed data

public ICommand LoadShapeFileCommand {
 get {
   if (loadShapeFileCommand == null) {
        // Use Lambda calculus to run the proper method when this command is applied
        loadShapeFileCommand = new GenericCommand(param => LoadShapeFile(), param => (mapViewModel != null));
   }
   return loadShapeFileCommand;
 }
}

The command patter used relies on Josh Smith’s relayCommand implementation (http://msdn.microsoft.com/en-us/magazine/dd419663.aspx)

The shape file is loaded as previously done, using SharpMap API. The map rendering is however now done in a different way using WPF rendering services instead of GDI+.
public void LoadShapeFile(string shapeFileName) {
      ...
      // The model has changed, notify the GUI of the change
      PropertyChangedEventArgs args = new PropertyChangedEventArgs("MapSnapshot");
      PropertyChanged(this, args);
}

The PropertyChanged event will end up in the main window.

        private void MapViewModelPropertyChanged(object sender, PropertyChangedEventArgs args) {
            string propertyName = args.PropertyName;
            if (propertyName == "MapSnapshot") {
                // Map snapshot has changed, reload it
                canvas1.Children.Clear();
                canvas1.ClipToBounds = true;
                canvas1.Children.Add(mapView.MapDrawing);
            }
        }
When doing this, watch out to properly setup the property name that has changed in the PropertyChangedEventArg. This is very important as it helps identify which part of the model did change (needs redrawing)
As you can see from the above code snapshot, the Canvas is updated by adding “mapView.Drawing” visual elements into its children.
public class MapView : FrameworkElement, INotifyPropertyChanged, IDisposable {
        private System.Windows.Media.VisualCollection children;
        ....
        public UIElement MapDrawing {
            get {
                return GetMapDrawing(mapViewModel.map);
            }
        }
        private UIElement GetMapDrawing(Map map) {
            children.Clear();
            VectorRenderer.map = map;
            foreach (SharpMap.Layers.ILayer layer in map.Layers) {
                // Render layer into layer drawing
                System.Windows.Media.DrawingVisual layerDrawing = VectorRenderer.Render(layer, map.Envelope);
                // Add rendered layer into drawing group
                AddVisualLayer(layerDrawing);
            }
            return this;
        }
        private void AddVisualLayer(System.Windows.Media.DrawingVisual visualLayer) {
            children.Add(visualLayer);
        }

As the rendering engine I’m working on these days is still “work in progress” only vector data can be rendered.

In this post, we have been through the main steps taken to put the MVVM pattern in action.

I also made some progress on writing my own WPF rendering engine which seems to do the work as expected. (see screenshot below)

Next, I’ll try to improve the rendering engine to support drawing other layer types, stay tuned…

Read Full Post »

WPF is making a lot of noise these days, I was wondering what this new technology is all about and decided to give it a try.

Coming from WinForm world, describing GUIs in text (XAML) seemed awkward to me, I struggled a bit  getting my first window up and running but managed to do so at the end.

In order to understand what is laying underneath WPF, I thought I should write my proper user control that can display something like a map.

Here is a screenshot of my first attempt to write such a control.

This control is based on SharpMap (http://sharpmap.codeplex.com/), reads and displays ESRI shape files.

For now, I am using SharpMap as it is. All what I did is to use the library, load some shape files and display them in an image control.

// Event handler for button click
public void OpenShapeFileClicked(object sender, EventArgs e){
      // ... ask the user to provide a shape file using a file open dialog control
      LoadShapeFile(path_provided_by_user);
}
public void LoadShapeFile(string fileName){
      SharpMap.Layers.VectorLayer myLayer = new SharpMap.Layers.VectorLayer("ShapeFiles");
      myLayer.DataSource = new SharpMap.Data.Providers.ShapeFile("path_to_file", true);
      myLayer.Style.Fill = System.Drawing.Brushes.GreenYellow;
      myLayer.Style.Outline = new System.Drawing.Pen(System.Drawing.Brushes.Gray);
      myLayer.Style.Line.Width = 2;
      myLayer.Style.Line = new System.Drawing.Pen(System.Drawing.Brushes.Black);
      myMap.Layers.Add(myLayer);
      myMap.ZoomToExtents();
      DrawImage();
}
private void DrawImage(){
      // Convert System.Drawing.Image to ImageSource to wich an image is bound
      using (Bitmap bitmap = new Bitmap(myMap.GetMap())){
             IntPtr hBitmap = bitmap.GetHbitmap();
             // Use System.Windows.Interop.Imaging.CreateBitmapSourceFromHBitmap
             BitmapSource bitmapSource = CreateBitmapSourceFromHBitmap(
                hBitmap,
                IntPtr.Zero,
                Int32Rect.Empty,
                BitmapSizeOptions.FromEmptyOptions());
             // Set image1 control data source
             image1.Source = bitmapSource;
      }
}

This project is not about rewriting a new SharpMap library for WPF but rather build up a WPF application from scratch based on SharpMap which I strongly believe will give me a great insight of the WPF technology and how GIS applications work.

In this little introduction, we demonstrated how easy it is to create a basic GIS application using WPF.

To be continued…

Read Full Post »