CLASS Services

The next version of CLASS extensions is going to be available by the end of summer with some really interesting goodies added. But, with HTML client (even before it’s official release) being in the center of interest, I feel like Silverlight extensions, are going to be soon, if not already, out of demand.
So, until I find the time to catch up with HTML client, I though on focusing more to stuff that are client implementation independent. That’s why, apart from the next version of CLASS Extensions for Silverlight clients, I plan on releasing a Server/Common side extension pack called CLASS Services.
If you are one of the blessed people that purchased CLASS Extensions, you already know of File Server service included, that provides file based content access to your server.
Two more of the services that are going to be included in this new CLASS Services pack can be found in this sample I posted in MSDN. Ok, I am not crazy, I know that if I share the code no-one will buy the extensions, but these two simple (as regards to the coding effort) services, are just a part of what CLASS Services is going to include. Either way, I believe my Lightswitch community contribution, even if not as extensive and steady as other people’s, makes clear that the idea is sharing and not making money out of it.
Regarding the sample, I must underline that the application itself is just an example. And I want to stress this point, because some would say that, what the sample application displays is not DI. The idea of the sample was not to display DI in action but just to display how you bind the RIA services together, to do whatever you like. MEF is used as a DI framework by the RIA Service to provide the required functionality.
Also, I have to admit that if properly implemented, by the book, restrictions and conventions in imports and exports of MEF (or any other DI framework for that matter) would make any additional configuration (like the one my sample suggests) pointless.
Nevertheless, IMHO, you really have to be a DI purist, not to recognize the potential provided by these two services. An example: In Computer Life, we plan (George stop calling me names ) to implement a generic (extension) way to support EAV model in our Lightswitch applications. It’s so obvious how useful the Domain Types RIA service will be for this implementation that I won’t even bother writing anything to prove it.
Anyway, I hope you find these two RIA services useful. I realize you have to go far beyond the “Coding is optional” motto to make good use of this kind of services, but I believe the Lightswitch development community, has already given the answer to that (not at all well-intentioned) “It’s the next MS-Access” hype…

Can I delete this?

In a previous post I suggested a way to override default LightSwitch add/edit behavior. In this post I will suggest an override to default delete behavior.
It has happened many times to delete entries, only to get an error message, when trying to save, informing me about reference constraints being violated and the action cannot be completed. This is not much of problem when you try to delete one record (apart from the fact this message is ugly and not one that I would allow my end users to deal with, for that matter). But when trying to delete many records you have no clue which was the one (or more) record(s) that cannot be deleted. In this case all you can do is delete-save until you find the record that cannot be deleted. And when you find it, you have to refresh your screen or undo the changes of the current record (a future article will suggest a way to do this).
What I needed was a good generic way to be able to set Delete_CanExecute result. This is my suggestion using (what else) extension methods, interfaces (Yann will love me for this ) and an attribute.
First the interface to help decide if an instance can be deleted or not:

public interface IDependencyCheck : IEntityObject
{
  bool CanDelete { get; }
}

Quite simple you must admit. Keep in mind interfaces you want to use in your common project have to be accessible both by Server and Client project also. If someone here thinks of EntityName_CanDelete, someone forgets that it is a server side hook.

Now an attribute to help us decide which of the dependencies an entity might have are strong ones, meaning that the referential integrity of the database or your business logic if you don’t use database constraints, would not allow an instance to be deleted.

[AttributeUsage(AttributeTargets.Class)]
public class StrongDependencyAttribute : Attribute
{
  public StrongDependencyAttribute(string dependencies) {
    this.dependencies = dependencies.ToStringArray();
  }

  public string[] Dependencies {
    get { return dependencies; }
  }

  private readonly string[] dependencies;
}

This attribute takes a list of strong reference property names as comma delimited string (or whatever delimiter you like for that matter) using these simple extensions:

public static string[] ToStringArray(this string strings) {
   return strings.ToStringArray(',');
}

public static string[] ToStringArray(this string strings, char delimiter) {
   return (from string item in strings.Split(new char[] { delimiter }, StringSplitOptions.RemoveEmptyEntries)
               select item.Trim()).ToArray();
}

If you want to use another delimiter just replace

this.dependencies = dependencies.ToStringArray();

in the constructor with (for example):

this.dependencies = dependencies.ToStringArray(';');

Or if you want, you can have one instance of the attribute for each property you want to check and avoid delimited strings altogether. Plenty of choices….

Keep in mind that if you declare no instance of this attribute but you implement IDependencyCheck then ALL dependencies will be considered as strong ones and checked for integrity.

This attribute needs to be accessible by the Common project only.

Now, all that said, there are two extension methods that will help us do the job.

The first one needs to be accessible by the Client project only:

public static bool CanDeleteSelection(this IScreenObject screen, string collectionName) {
   if (!screen.HasSelection(collectionName))
     return false;
   IVisualCollection collection = screen.Details.Properties[collectionName].Value as IVisualCollection;
   if (collection.SelectedItem is IDependencyCheck)
     return (collection.SelectedItem as IDependencyCheck).CanDelete;
   return true;
}

Please note I am using IScreenObject.HasSelection extension that has already been introduced in a previous post. That’s why I am not checking if the cast to IVisualCollection is successful (not null).

The second one has to be accessible by the Common project only:

    public static bool CanDelete(this IEntityObject entity) {
      IEnumerable<IEntityCollectionProperty> collectionProperties = 
        entity.Details.Properties.All()
        .Where(p => p.GetType().GetInterfaces()
          .Where(t => t.Name.Equals("IEntityCollectionProperty"))
          .FirstOrDefault() != null)
        .Cast<IEntityCollectionProperty>();
      if (collectionProperties == null)
        return true;
      List<string> strongDependencies = new List<string>();
      IEnumerable<StrongDependencyAttribute> dependencies = 
        entity.GetType()
        .GetCustomAttributes(typeof(StrongDependencyAttribute), false)
        .Cast<StrongDependencyAttribute>();
      foreach (StrongDependencyAttribute dependency in dependencies)
        strongDependencies.AddRange(dependency.Dependencies);
      bool hasDependencies = strongDependencies.FirstOrDefault() != null;
      bool canDelete = true;
      foreach (IEntityCollectionProperty property in collectionProperties) {
        if (hasDependencies &&strongDependencies.FirstOrDefault(d => d.Equals(property.Name)) == null)
          continue;
        IEnumerable value = entity.GetType()
          .GetProperty(string.Format("{0}Query", property.Name))
          .GetValue(entity, null) as IEnumerable;
        try {
          if (value != null && value.GetEnumerator().MoveNext()) {
            canDelete = false;
            break;
          }
        }
        catch {
          continue;
        }
      }
      return canDelete;
    }

Although it’s obvious at first glance what the code does , I will give brief explanation:

If there is not a StrongDependencyAtttibute defined for the entity then all reference properties are checked and if at least one has members then the entity cannot be deleted. If a StrongDependencyAtttibute is defined for the entity then only reference properties included in the attribute are checked. That’s all…

If you manage to read the code (I am not very proud about the absence of comments) you will notice that only one-to-many and many-to-many references are handled. In my world one-to-one references mean inheritance and in this case both objects should be deleted. But what if the base object can be deleted (has no direct references) and the derived object has? Again in my world, if you are trying to delete the base object you are already doing it wrong! Anyway if someone lives in a world other than mine (I am very democratic guy ) and wants to support one-to-one relations all he/she has to do is find where IEntityCollectionProperty definition is and look for the respective property type (I believe it is IEntityReferenceProperty but I am not quite sure).

And for the end an example so that anyone can see what all of the above end up to:

Suppose you have a Customer entity. And this Customer entity has a collection of Orders. The property of the Customer entity that holds these orders is called CustomerOrders. In your Datasource you right-click Customer entity and select View Table Code. A partial class implementation file is created (if it does not already exist). Modify the definition of you class as follows:

[StrongDependency("CustomerOrders")]
public partial class Customer : IDependencyCheck {
...

  #region IDependencyCheck members 
  public bool CanDelete {
    get { return this.CanDelete(); }
  }
  #endregion IDependencyCheck members
}

Remember to reference (using) the namespace where your extension method (CanDelete) is declared.

Please note that IDependencyCheck gives you the potential to write whatever else hardcoded (or not) check you want in your CanDelete property implementation. In the code above I just call the extension method I introduced earlier. But you can do whatever you want. You can even skip the dependencies mechanism suggested altogether. The client side extension will still work.

So in the screen that you have your list of Customers right click the Delete command of the list or grid and in the CustomersDelete_CanExecute just write:

partial void CustomersDelete_CanExecute(ref bool result){
  result = this.CanDeleteSelection("Customers");
}

As partial implementation of Execute write:

partial void CustomersDelete_Execute(){
  this.Customers.DeleteSelected();
}

I know some of you have already noticed the overhead of potentially loading the dependent objects the first time you select an item of your list or grid. I cannot argue with that, except for the fact that my approach is suggested for intranet implementations (I am not sure I would do something like that over the web) and the fact that the time under these circumstances is an acceptable price to pay in order to avoid the annoying referential integrity message. At least in my world . 

De-fault Screens

If you have tried implementing something that required a mechanism more complex than the default one provided by LightSwitch regarding default screen selection, you already are familiar with the limitations.
For the ones that don’t get what I mean, a very common example is open the details screen of an object from a list containing items from a view (database view) that include the id of the original object. Imagine you have a Customer entity and a CustomerInfo entity that comes from a database view bringing aggregated data of the customer from various tables. Selecting and editing the CustomerInfo you want to edit the Customer not the CustomerInfo.
The above is only one fast example. Another would be having Add/Edit Screens i.e. screens that can handle both create or edit an existing object. In any case if you need to modify the default signature of a default screen (adding an extra parameter for example) causes this screen to be unselected as default screen because LightSwitch mechanism for default screens cannot handle it.
Going back to the Customer/CustomerInfo example, I have come across at least a couple of posts in threads trying to find a way to “override” the default click behavior of links in collections. In our example clicking on the link of the customer’s name (for example) will open a default screen for CustomerInfo.
So, one easy solution is creating a default screen for CustomerInfo where upon load you will open the Customers default screen passing the id of the customer and close the current one. But how do you handle allowing only one instance of a details screen per object? Or all other issues that might come come up? (Imagine you have more than one view entities for the Customer entity).
Anyway, this is my approach at solving this issue, that works fine for me in a big project whose development runs for almost a year now (and by the way is the reason for not posting very often –understatement of the year- here lately):
Reintroducing the ShowDefaultScreen method of the Application in the Client project along with an interface defined and implemented by all screens that need to be involved.

• This is the interface definition:

public interface IDefaultScreen : IScreenObject
{
    bool Shows(IEntityObject instance);
    IEntityObject Object { get; }
}

• This is the code in Application.cs:

public new void ShowDefaultScreen(IEntityObject instance) {
      this.ShowOrFocusDefaultScreen(instance);
}

public void ShowOrFocusDefaultScreen(IEntityObject entity) {
  foreach (IActiveScreen activeScreen in this.ActiveScreens) {
    if ((activeScreen.Screen is IDefaultScreen) &&
        (activeScreen.Screen as IDefaultScreen).Shows(entity)) {
          activeScreen.Activate();
          return;
        }
      }
  entity.Details.Dispatcher.BeginInvoke(() => {
    if (!HandleEntityDefaultScreen(entity))
      base.ShowDefaultScreen(entity);
  });
}

private bool HandleEntityDefaultScreen(IEntityObject entity) {
  switch (entity.GetType().Name) {
    case "CustomerInfo": {
        this.ShowEditCustomer((entity as CustomerInfo).Id);
        return true;
      }
      .
      .
      .
  }
  return false;
}

In the code displayed above the implementation in HandleEntityDefaultScreen code is just to demonstrate the concept. The ellipsis below the first case implies you can write whatever is required by your application.

• And finally this is the code from the Customer details screen that implements IDefaultScreen:

public partial class EditCustomer : IDefaultScreen
{
   .
   .
   .
  #region IDefaultScreen Members
  public bool Shows(IEntityObject instance) {
    return (instance is CustomerInfo) &&
           (instance as CustomerInfo).Id.Equals(this.CustomerProperty.Id);
  }

  public IEntityObject Object {
    get { return this.CustomerProperty; }
  }
  #endregion
}

The careful reader will notice that the IEntityObject Object { get; } property exposed by IDefaultScreen is not used in the sample code. I ported the code exactly as I use it in my projects (apart from the changes made to demonstrate the Customer/CustomerInfo paradigm) where this property is used for a different purpose. You can either remove it or find a good use for it .

Just Visiting or (not so) Simple Extension Methods (part 4)

One of my favorite design patterns since the age of c++ is the Visitor Pattern. I will not explain here what the visitor pattern is. But, if you know how to use the design pattern this is a post worth reading.
One may ask what the visitor pattern has to do with LightSwitch. Well, it doesn’t! I mean not exclusively. But the code below provides a full visitor pattern implementation background that can also be used in LS. Also, a part of the implementation is ideal for LS, since one of the challenges I had to face, when trying to widely use the visitor pattern, was make it work for “sealed” classes, classes that were not written by me and could not easily fit to my –reflection based- visitor pattern implementation. To solve this the best thing I could think of was “wrapping”. And working with LS, most of the classes (apart from the ones that belong to the domain/datasource) are actually “sealed” in the way described above.(I have to note here that this reflection-based implementation is a revision of an implementation I found (if I recall correct) in CodeProject).
First the two basic interfaces that have to be defined:
The IVisitor interface that has to be implemented by the “worker” (or helper to follow the VB-oriented naming of the LS Team ) class.

public interface IVisitor
{
  void VisitDefault(IVisitorTarget source);
}

Then the IVisitorTarget interface (implied above) that has to be implemented by the class to be “consumed/visited” by the Visitor.

public interface IVisitorTarget
{
  void Accept(IVisitor visitor);
}

Ok, I know, nothing much up to now. But as I say “The longest journey starts with the first step”…(yes I am an old Chinese philosopher)

To have your visitor pattern in place you need an extension method to do the trick:

public static bool ConcreteVisit(this IVisitorTarget target, IVisitor visitor) {
  Type[] types = new Type[] { target.GetType() };
  MethodInfo mi = visitor.GetType().GetMethod("Visit", types);

  if (mi == null)
    return false;

  mi.Invoke(visitor, new object[] { target });
  return true;
}

Now that these pieces are in place let’s say we have a class called DisplayNameBuilder that implements the IVisitor interface. Also, let’s say we have a Customer entity in our datasource and we want it to implement the IVisitorTarget interface. All we have to do is open Customer entity in the designer, click Write Code and change the class declaration to:

public partial class Model : IVisitorTarget

make sure you are using the namespace of the class where the above extension method is implemented and implement the IVisitorTarget interface like this:

#region IVisitorTarget Members
public void Accept(IVisitor visitor) {
  if (!this.ConcreteVisit(visitor))
    visitor.VisitDefault(this);
}
#endregion

Also this is a sample implementation of the DisplayNameBuilder class:

public class DisplayNameBuilder : IVisitor
{
  public string Name{
    get;
    private set;
  }

  public void VisitDefault(IVisitorTarget visitorTarget){
    Name = visitorTarget.ToString();
  }

  public void Visit(Customer visitorTarget){
    Name = string.Format("{0}{1}{2}", visitorTarget.FirstName, string.IsNullOrWhiteSpace(visitorTarget.FirstName) ? "", " ", visitorTarget.LastName);
  }
}

In the above code please note these:

1. The code was written in Live Writer as it’s demo code, so maybe it does not compile as is.


2. The customer is implied to have a nullable FirstName property and a not nullable LastName (which I believe it’s a fair assumption and I agree with me).


3. The implementation, as obvious, is domain aware as it knows what Customer is. This implies that the ideal place for this class to live is in the Common project.

Now lets say you create a calculated string field in Customer entity called hmmmmm… DisplayName (surprised? It’s a gift I have regarding giving original names). This would be the code that would implement the calculation of DisplayName property:

partial void DisplayName_Compute(ref string result) {
  DisplayNameBuilder builder = new DisplayNameBuilder();
  this.Accept(builder);
  result = builder.Name;
}

I HAVE to stress one more time that the code is for demonstration purposes only. It’s not just you, it IS too much fuss for nothing.

Ok, now it should be easier to understand the problem of using the visitor pattern with pre-defined classes. You cannot add the IVisitorTarget behavior to most of the classes, automatically generated by LS.

So this is the solution to the problem.

public class IVisitorWrapper<TObjectType> : IVisitorTarget
{
  public TObjectType Content {
    get;
    set;
  }

  #region IVisitorTarget Members
  public void Accept(IVisitor visitor) {
    if (!this.ConcreteVisit(this, visitor))
      if (!this.VisitWrapper(this, visitor))
        visitor.VisitDefault(this);
  }
  #endregion
}

This is a wrapper class (one could say it is an IVisitorTarget decorator). If you are still with me and you keep on reading the code you have already noticed that VisitWrapper extension method is not yet defined, so this is it:

public static bool VisitWrapper<TObjectType>(this IVisitorWrapper<TObjectType> wrapper, IVisitor visitor) {
  Type[] types = new Type[] { typeof(TObjectType) };
  MethodInfo mi = visitor.GetType().GetMethod("Visit", types);

  if (mi == null)
    return false;

  mi.Invoke(visitor, new object[] { wrapper.Content });
  return true;
}

Now you can wrap any type of object and use the DisplayNameBuilder to build it’s display name. Also (as –I hope- you guessed) you have to implement the respective Visit method in DisplayNameBuilder in order not go get the type name back as Name.

Well, as Porky says: That’s all Folks . I hope that except for exhausting it was also interesting.

Application Logo (tailored on customer demand)

This is the most recent comment in the Application Logo Sample post:
“Hi Kostas,
I wanted to find out if there is a way of opening an active screen as soon as I click on the logo. I have tried but my attempt was in vain. If it is possible, could you please help me out?
Thanks,
Darryn”
My first reaction was to ask for details (typical way of buying time ). But then I though I could do the changes required and post a new article. So simply put, this is the answer I propose to Darryn:

public static void AddLogo(this Microsoft.LightSwitch.Client.IClientApplication application, System.Windows.HorizontalAlignment alignment, string screenName) {
  Microsoft.LightSwitch.Threading.Dispatchers.Main.BeginInvoke(() => { LogoPlacement.AddLogo(application, System.Windows.Application.Current.RootVisual, alignment, screenName); });
}


private static class LogoPlacement
{
  internal static void AddLogo(Microsoft.LightSwitch.Client.IClientApplication application, UIElement element, System.Windows.HorizontalAlignment alignment, string screenName) {
    if (rcb != null) return;

    for (int i = 0; i < System.Windows.Media.VisualTreeHelper.GetChildrenCount(element); i++) {
      if (rcb != null)
        return;
      UIElement child = (UIElement)System.Windows.Media.VisualTreeHelper.GetChild(element, i);
      AddLogo(application, child, alignment, screenName);
    }
    if (element is Microsoft.LightSwitch.Runtime.Shell.Implementation.Standard.RibbonCommandBar) {
      rcb = element as Microsoft.LightSwitch.Runtime.Shell.Implementation.Standard.RibbonCommandBar;
      Image myImage = new Image() {
        Stretch = System.Windows.Media.Stretch.Uniform,
        Margin = new Thickness(2, 8, 14, 8),
        HorizontalAlignment = alignment,
        Cursor = System.Windows.Input.Cursors.Hand
      };
      myImage.SetValue(ComponentViewModelService.ViewModelNameProperty, "Default.LogoViewModel");
      myImage.SetBinding(Image.SourceProperty, new System.Windows.Data.Binding { Path = new PropertyPath("Logo") });
      if (!string.IsNullOrWhiteSpace(screenName))
        myImage.MouseLeftButtonUp += (s, e) => { application.Details.Dispatcher.BeginInvoke(() => application.Details.Methods[string.Format("Show{0}", screenName)].CreateInvocation().Execute()); };
      myImage.SizeChanged += (s, e) => {
        double left = (s as Image).HorizontalAlignment == HorizontalAlignment.Left ? e.NewSize.Width + 10.0 : 0.0;
        double right = (s as Image).HorizontalAlignment == HorizontalAlignment.Right ? e.NewSize.Width + 10.0 : 0.0;
        rcb.Padding = new Thickness(left, 0, right, 0);
      };
      ((Grid)rcb.Parent).Children.Add(myImage);
    }
  }

With bold I have marked the additions/changes I had to make to accommodate Darryn’s request.
I hope Darryn is happy .

Simple Extension Methods (part 3)

One very common task when one is implementing the business logic in Common project is writing custom PropertyName_Changed partial methods.
This “approach” has an undesired “side-effect” when the property is an entity property (as compared to screen properties). The property changes many times during the lifetime of the object. While loading, while deleting etc., whereas what we want to implement is handle the “actual” property changes caused either by the user or by some other piece of business logic.
I found myself debugging code that shouldn’t be executing (according to my logic, obviously LS had another opinion) before ending up writing a small extension method to use in all PropertyName_Changed partial methods. I don’t claim this method will help you avoid ALL unwanted business logic code execution, but covers most of the cases. I would be glad to read any comments improving it in a generic way.
Here is the code of the extension method:

public static bool IgnorePropertyChange(this IEntityObject entity) {
  return entity.Details.EntityState == EntityState.Unchanged ||
         entity.Details.EntityState == EntityState.Discarded ||
         entity.Details.EntityState == EntityState.Deleted;
}

And here how all my property changed partial implementations for entity properties look like

partial void ScheduledEndDate_Changed() {
  if (this.IgnorePropertyChange())
    return;
  .
  .
  .
}

This looks (and is) simple but the concept, alone, of “unwanted” code execution is important and should be taken good care of. If not using this extension method any other way one may like .

Simple Extension Methods (part 2)

In the previous post I presented an extension method used mostly for overriding the edit and delete commands of a collection. One may ask “why do I want to do this?”. Apart from any other requirements/business logic dependent reason one might want to implement, for me there is one simple yet important reason: I don’t like at all (to be kind) the default add/edit modal windows when adding or editing an entry. It’s not a coincidence that the FIRST sample I wrote for LightSwitch and posted in the Samples of msdn.com/lightswitch was a set of extension methods and contracts to easily replace standard modal windows with custom ones.
Most of the times when I have an editable grid screen, selecting Add or Edit I DON’T want the modal window to pop-up, I just want to edit in the grid. Or in list and details screen I want to edit the new or existing entry in the detail part of the screen.
This is the main reason I most of the times override the default Add/Edit command behavior. And for this reason I created and use the next two extension methods:

public static void AddFocus<T>(this VisualCollection<T> collection, string focusControl, bool inCollection)
      where T : class, IEntityObject {
      collection.AddNew();
      collection.EditFocus(focusControl, inCollection);
    }

    public static void EditFocus<T>(this VisualCollection<T> collection, string focusControl, bool inCollection)
      where T : class, IEntityObject {
      if (focusControl != null) {
        try {
          IContentItemProxy control = null;
          if (inCollection)
            control = collection.Screen.FindControlInCollection(focusControl, collection.SelectedItem);
          else
            control = collection.Screen.FindControl(focusControl);
          Microsoft.LightSwitch.Threading.Dispatchers.Main.BeginInvoke(() => {
            try {
              control.Focus();
            }
            catch {
              collection.EditSelected();
            }
          });
        }
        catch {
          if (collection.HasSelection())
            collection.EditSelected();
        }
      }
    }

So what you have to do in your MyCollectionAddAndEditNew_Execute partial method is write something like:

MyCollection.AddFocus("TheNameOfTheControlYouWantToFocus", true);

or for MyCollectionEdit_Execute partial method:

MyCollection.EditFocus("TheNameOfTheControlYouWantToFocus", true);

if you have a list detail screen the same code applies but with inCollection = false.

One could implement one method for collections and one for details. Also notice where constraints applied to T generic parameter. The class constraint is required for declaring VisualCollection<T>. The IEntityObject constraint is required so that we avoid the arbitrary casting of collection.SelectedItem to IEntityObject so as to be passed to FindControlInCollection as second parameter.

Notice that in the extension methods I don’t check if the control name exists. I just use try catch. On error (if the control name is not valid) I choose to fall-back to the default behavior. You can change this without any intellectual rights being violated .