Archive

Be seated safely before reading on. This kicks ass.

I have already written enough abstract words in previous posts; let's now introduce PostSharp 2.0 on a real example: the implementation of  NotifyPropertyChanged, one of the most frequently used patterns for all adepts of MVC designs.

It's not just about implementing the INotifyPropertyChanged interface; it's also about modifying every property setter. It's deadly simple and deadly boring. And gets you more than one step away from the idea you have of aesthetical code.

Enough words. See the implementation using PostSharp 2.0. There's a lot of new concepts out there, so let's start easily:

[Serializable]
[IntroduceInterface( typeof(INotifyPropertyChanged))]
public sealed class NotifyPropertyChangedAttribute : 
   InstanceLevelAspect, INotifyPropertyChanged
{
    public void OnPropertyChanged( string propertyName )
    {
        if ( this.PropertyChanged != null )
        {
            this.PropertyChanged( this.Instance, 
               new PropertyChangedEventArgs( propertyName ) );
        }
    }

    [IntroduceMember]
    public event PropertyChangedEventHandler PropertyChanged;

    [OnLocationSetHandler, 
     MulticastSelector( Targets = MulticastTargets.Property )]
    public void OnPropertySet( LocationInterceptionArgs args )
    {
        if ( args.Value == args.GetCurrentValue() ) return;

        this.OnPropertyChanged( args.Location.PropertyInfo.Name );
    }
}

The aspect is used this way:

[NotifyPropertyChanged]
public class TestBaseClass
{
    public string PropertyA { get; set; }
}

Some words of explanation.

First, look at introductions: IntroduceInterface tells that the interface INotifyPropertyChanged must be introduced into the target type. And where is the interface implemented? Well, by the aspect itself! Note that PropertyChanged is itself introduced to the target type as a public member thanks to the IntroduceMember custom attribute (without this attribute, the interface would be implemented explicitely).

The aspect actually becomes an extension of the class; aspect instances have the same lifetime as objects of the class to which the aspect is applied... just because the aspect inherits from InstanceLevelAspect.

Got it? So let's continue to the OnPropertySet method. It is marked by custom attribute OnLocationSetHandler: it makes this method a handler that intercepts all calls to the setter of all target properties. And which are target properties? This is told by custom attribute MulticastSelector: all properties in this case, but we could filter on name, visibility and all features you are used to. We could have used the same handler to handle access to properties (it would have turned the field to a property as a side effect).

Did you get it? You can now catch accesses to properties or fields using the same aspects. Same semantics, same aspect. Simple, powerful.

Now look at the body of method OnPropertySet and see how easy it is to read the current value of the property.

The code above works on an isolated class, but a class is rarely isolated, right? More of the time, it derives from an ancestor and have descendants. What if the ancestor already implements interface INotifyPropertyChanged? The code above would trigger a build time error. But we can improve it. What do we want, by the way? Well, if the class above already implements INotifyPropertyChanged, it must also have implemented a protected method OnPropertyChanged(string), and we have to invoke this method. If not, we define this method ourselves. In both cases, we need invoke this method from all property setters.

Let's turn it into code:

[Serializable]
[IntroduceInterface( typeof(INotifyPropertyChanged), 
                     OverrideAction = InterfaceOverrideAction.Ignore )]
[MulticastAttributeUsage( MulticastTargets.Class, 
                          Inheritance = MulticastInheritance.Strict )]
public sealed class NotifyPropertyChangedAttribute : 
   InstanceLevelAspect, INotifyPropertyChanged
{
   
    [ImportMember( "OnPropertyChanged", IsRequired = false )] 
    public Action<string> BaseOnPropertyChanged;

    [IntroduceMember( Visibility = Visibility.Family, 
                      IsVirtual = true, 
                      OverrideAction = MemberOverrideAction.Ignore )]
    public void OnPropertyChanged( string propertyName )
    {
        if ( this.PropertyChanged != null )
        {
            this.PropertyChanged( this.Instance, 
               new PropertyChangedEventArgs( propertyName ) );
        }
    }

    [IntroduceMember( OverrideAction = MemberOverrideAction.Ignore )]
    public event PropertyChangedEventHandler PropertyChanged;

    [OnLocationSetHandler, 
     MulticastSelector( Targets = MulticastTargets.Property )]
    public void OnPropertySet( LocationInterceptionArgs args )
    {
        if ( args.Value == args.GetCurrentValue() ) return;

        if ( this.BaseOnPropertyChanged != null )
        {
            this.BaseOnPropertyChanged( args.Location.PropertyInfo.Name );
        }
        else
        {
            this.OnPropertyChanged( args.Location.PropertyInfo.Name );
        }
    }
}

This time, the code is complete. No pedagogical simplification. Look at custom attributes IntroduceInterface and IntroduceMember: I have added an OverrideAction; it tells that interface and member introductions will be silently ignored if already implemented above.

Now look at field BasePropertyChanged: its type is Action<string> (a delegate with signature (string):void) and is annotated with custom attribute ImportMember. At runtime, this field with be bound to method OnPropertyChanged of the base type. If there is no such method in the base type, the field will simply be null. So, in method OnPropertySet, we can now choose: if there was already a method OnPropertyChanged, we invoke the existing one. Otherwise, we invoke the one we are introducing.

Thanks to ImportMember, we know how to extend a class that already implement the pattern. But how to make our own implementation extensible by derived classes? We have to introduce the method OnPropertyChanged and make it virtual. It's done, again, by custom attribute IntroduceMember.

That's all. You can now use the code on class hierarchies, like here:

[NotifyPropertyChanged]
public class TestBaseClass
{
    public string PropertyA { get; set; }
}
public class TestDerivedClass : TestBaseClass
{
    public int PropertyB { get; set; }
}
class Program
{
    static void Main(string[] args)
    {
        TestDerivedClass c = new TestDerivedClass();
        Post.Cast<TestDerivedClass, INotifyPropertyChanged>( c ).PropertyChanged += OnPropertyChanged;
        c.PropertyA = "Hello";
        c.PropertyB = 5;
    }

    private static void OnPropertyChanged( object sender, PropertyChangedEventArgs e )
    {
        Console.WriteLine("Property changed: {0}", e.PropertyName);
    }
}

With PostSharp 1.5, you could do implement easy aspects easily. With PostSharp 2.0, you can implement more complex design patterns, and it's still easy.

Happy PostSharping!

-gael

P.S. Kicking?

Yesterday, C# MVP Job Skeet blogged about a  way to add generic constraints for enums and delegates. He noticed that, while the C# compiler forbids such constructions, it is perfectly possibly in MSIL. And he is fully right with this.

Jon uses a post-build step to add modify generic constraints. His process: ILDASM, then find-and-replace, then ILASM. That’s possible because the requirement is pretty easy.

But here he challenged me:

I've looked at PostSharp and CCI; both look way more complicated than the above.

Sure, PostSharp would be more complex, but Jon’s ILASM-based solution works only in best cases. There are plenty of things that you have to keep in mind while building a post-build step that modifies MSIL. For instance, how will ILASM find referenced assemblies? What with signed assemblies?  Sure, you can find a solution to everything – after all, PostSharp uses ILASM as a back-end (contrarily to rumors, it does not use ILDASM). But, believe me, addressing these side issues will prove more difficult than the core job itself.

And anyway, why would it be so complex to implement these rather simple requirements using PostSharp? PostSharp is designed so that easy jobs get easily done.

I took the challenge and implemented an add-in adding arbitrary constraints to generic parameters. I measured time. It took me 30 minutes. NDepend counts 12 lines of code.

It’s based on custom attributes. PostSharp loves custom attributes. I defined a custom attribute AddGenericConstraintAttribute that can be applied to generic parameters. Here is how it’s used:

public static class EnumHelper
{
    public static T[] GetValues<[AddGenericConstraint( typeof(Enum) )] T>() 
       where T : struct
    {
        return (T[]) Enum.GetValues( typeof(T) );
    }
}

The project can be downloaded from http://postsharp-user-samples.googlecode.com/files/AddGenericConstraint.zip. In order to use the plug-in, follow the documentation.

Here are the principal steps I followed to create this plug-in:

1.       Create a new solution AddGenericConstraint.sln.

2.       Create a new class library project AddGenericConstraint.csproj. This project will be the public interface of the plug-in.

3.       Create the class AddGenericConstraintAttribute.cs:

[AttributeUsage( AttributeTargets.GenericParameter )]
public sealed class AddGenericConstraintAttribute : Attribute
{
    public AddGenericConstraintAttribute( Type type )
    {
    }
}

4.       Create a new class library project AddGenericConstraint.Impl.csproj. This will contain the plug-in implementation.

5.       Add references to PostSharp.Public.dll, PostSharp.Core.dll, AddGenericConstraint.csproj.

6.       Create a class AddGenericConstraintTask. This is the real implementation class. This is the hard point.

public class AddGenericConstraintTask : Task
{
    public override bool Execute()
    {
        // Get the type AddGenericConstraintAttribute.
        ITypeSignature addGenericConstraintAttributeType = 
            this.Project.Module.FindType(
               typeof(AddGenericConstraintAttribute),
               BindingOptions.OnlyDefinition | BindingOptions.DontThrowException );

        if ( addGenericConstraintAttributeType == null )
        {
            // The type is not referenced in the current module. There cannot be a custom attribute
            // of this type, so we are done.
            return true;
        }

        // Enumerate custom attributes of type AddGenericConstraintAttribute.
        AnnotationRepositoryTask annotationRepository = 
            AnnotationRepositoryTask.GetTask( this.Project );

        IEnumerator customAttributesEnumerator =
            annotationRepository.GetAnnotationsOfType( 
              addGenericConstraintAttributeType.GetTypeDefinition(), false );

        while ( customAttributesEnumerator.MoveNext() )
        {
            // Get the target of the custom attribute.
            GenericParameterDeclaration target = (GenericParameterDeclaration)
                customAttributesEnumerator.Current.TargetElement;
 
            // Get the value of the parameter of the custom attribute constructor.
            ITypeSignature constraintType = (ITypeSignature)
                customAttributesEnumerator.Current.Value.
                 ConstructorArguments[0].Value.Value;

            // Add a generic constraint.
            target.Constraints.Add( constraintType );

            // Remove the custom attribute.
            ((CustomAttributeDeclaration) customAttributesEnumerator.Current).Remove();
        }

        return base.Execute();
    }
}

7.       Add an XML file AddGenericConstraint.psplugin to the project. This file will describe your plug-in. In file properties, specify “Copy to Output Directory: Copy Always”.

<?xml version="1.0" encoding="utf-8" ?>
<PlugIn xmlns="http://schemas.postsharp.org/1.0/configuration">
  <TaskType Name="AddGenericConstraint" Implementation="AddGenericConstraint.Impl.AddGenericConstraintTask, AddGenericConstraint.Impl" Phase="Transform">
    <Dependency TaskType="Remove" Position="After"/>
    <Dependency TaskType="AnnotationRepository"/>
  </TaskType>
</PlugIn>

8.       Go back to project AddGenericConstraint. Add a reference to PostSharp.Public.dll.  Add the following on the top of the class AddGenericConstraintAttribute to bind the custom attribute to its implementation:

[RequirePostSharp( "AddGenericConstraint", "AddGenericConstraint" )]

9.       Create a test project. Add references to PostSharp.Public.dll and AddGenericConstraint.csproj. In project properties, add directory “..AddGenericConstraint.ImplinDebug” to reference paths.

10.   You are done.

What’s the catch? Maybe that, if you don’t know PostSharp Core and MSIL, it will take you days to come with the 12 lines of code that form the implementation. But when you got the knowledge, you are incredibly productive. And all the issues caused by the integration in the build process are solved for you. Believe me, after 5 years of existence, there is a huge knowledge behind PostSharp integration.

Happy PostSharping!

-gael

PS. It took me longer to write this blog than the implementation itself.

 

Before starting to blog about new features of PostSharp 2.0, it's good to remind the design objectives of this release.

Stated in one sentence: PostSharp is the leading aspect framework for .NET and will remain.

This was already true with PostSharp 1.5. In next version, the distance between PostSharp and other frameworks will be huge. You will judge by yourself in next posts.

Backward compatibility

As everyone knows, being the leader comes with greater responsibilities. And the responsibilities that lay on the shoulders of a framework are harsh. People working in Microsoft or Sun could easily confirm it: once your framework starts being widely used, maintaining backward compatibility is not optional. Every design decision will have consequences years later. And design errors can have catastrophic consequences for the community.

A framework differs from a library in the fact that a library is intended to be used, whereas the framework is intended to be extended. Designing a framework that is both highly extensible and prepared for backward compatibility is very challenging. As the framework developer, we have to be very specific about our supported extension points, and make sure that additional functionalities can be exposed through these extension points without breaking binary compatibility.  

PostSharp 1.5 was designed to be "approximately backward compatible" at source level. It took the assumption that, as the aspect developer, you can spend a few minutes making minor fixes in your source code when you upgrade. However, this works only because you are the developer of the only aspects used on your computer.

Now, what if you get aspects from other vendors? Say you are using DataObjects.NET 4.0 (DO). It comes with its own PostSharp aspects. Can you use upgrade PostSharp at will? No, you will have to use the version DO was compiled for. Now what if you use also aspects from Log4PostSharp? What if two vendors compile using different versions of PostSharp? Most probably, it would break. As you can see, approximate source-level compatibility is not enough when you develop a framework that can load plug-ins from multiple vendors (something that Linux kernel developers have never understood).

Being the leading aspect framework for .NET, many vendors will want to deliver aspects for their own products - based on PostSharp. So these vendors will have the possibility to redistribute PostSharp with their own products, and get very attractive commercial conditions to do so (much more attractive to small companies than for PostSharp 1.5). But what if a user upgrades PostSharp after having installed a third-party plug-in? The plug-in would be executed in a version of PostSharp that is more recent than the one it has been designed for. But it will have to work! And it will, because it has been engineered to.

Dependency Solving

Backward compatibility is just one requirement brought by the multiple vendor scenario.

Another issue is dependency solving. Say that vendor A delivers a caching aspect and vendor B an authentication aspect. Clearly, authentication should always be checked before the cache look up is performed. How to express this dependency if A and B don't know about each other? In PostSharp 1.5, dependencies are solved by setting the priority of aspects using an integer value. This works right when the team is in charge of all aspects, but it clearly does not scale to larger teams or multi-vendor scenarios. So PostSharp 2.0 comes with a brand new symbolic dependency engine based on topological sort. The dependency engine is able to detect conflicts between aspects and dependency cycles. This engine is actually at the heart of the new architecture.

Large Vision of AOP

Because it needs to be backward compatible and extensible in the future, PostSharp 2.0 also needs to implement a large vision of aspect-oriented programming.  PostSharp 1.5 implemented a rather small subset of the features offered by AspectJ. It would not have been possible to implement AspectJ based on PostSharp Laos. Even if PostSharp 2.0 does not implement all the features of AspectJ, I am quite confident that it would be possible in to do so.

PostSharp 2.0 will already support method interception, property interception, field interception, event interception, property introduction, event introduction, method introduction, interface introduction, handlers (aka advises in AspectJ), selectors (aka pointcuts in AspectJ). No heavy workaround, like CompoundAspect in PostSharp 1.5, will be necessary to implement a pattern like NotifyPropertyChanged. A dozen of clean lines of code will do the job. I will blog about this very soon.

Developer's Productivity

As teams start using aspects in large projects, it is sometimes difficult to answer two questions:

  • Which aspects are being applied to the method I am now looking at?
  • Conversely, which methods is a given aspect applied to?

With the new Visual Studio add-in, you won't even have to click to have the answers. Would a code element be enhanced by an aspect, it will be lightly underlined. The tooltip text: the list of aspects with hyperlinks.

The objective is simple: stop experimenting with aspects, start being productive with them. And get support from the IDE.

Runtime Optimization

Last but not least, PostSharp 2.0 was designed to emit optimized instructions so that the overhead of aspects, with respect to hand-written code, is as small as possible. Generally, extensibility is achieved at the cost of performance. Here, not. And, believe me, it was not easy to program it. But I did, and it works wonderfully.

What's not there?

Some say that the first commercial version of a product should deliver the smallest possible set of features that still make it interesting to early adopters to buy. PostSharp 2.0 is clearly beyond this line. However, there are things we could have improved but did not fit in the 2.0 schedule: build performance is the principal. We did already big efforts in PostSharp 1.5, but 2.0 will not perform better. We have ideas how to improve it and will work on it in future releases. If you are experiencing performance problems with PostSharp 1.5, be sure that it is installed on your computer using the installer (not the zip package) so that native images are used. PostSharp startup time will be significantly reduced (and PostSharp 1.5/2.0 starts as many times as you have projects in the solution).

What's taken over from PostSharp 1.5

There are basically two components in PostSharp 1.5: PostSharp.Core and PostSharp.Laos.Weaver. PostSharp.Core is kept without significant change (so, if you are just using this components, upgrading to 2.0 will be minor work). However, PostSharp.Laos.Weaver is dropped and replaced by the new aspect framework I talked about before.

PostSharp.Laos itself (the layer most people use directly) will have partial source-level compatibility to version 1.5. Actually, PostSharp.Laos will run emulation on the top of the new aspect framework. Most unit tests done for PostSharp Laos 1.5 still run in version 2.0 in emulation mode, but not all. The emulation assembly is fully annotated by [Obsolete] custom attributes to help you migrating to the new framework.

Sure that it will hurt some people. But we had to do it. The objective of PostSharp Laos was to be a quick win in Pareto meaning: 20% of investment on the top of PostSharp Core, covering 80% of use cases (at least these perceived at design time). The new framework was much more expensive and will cover more use cases. But PostSharp Laos, especially the backend PostSharp.Laos.Weaver, was never engineered to be really extensible. So, yes, we fully break compatibility of PostSharp.Laos.Weaver, but it was the price to have compatibility with the new framework.

Summary

I have chatted a lot about the vision behind PostSharp 2.0. I wrote a lot of words because I wanted to express the context and the challenges we are trying to solve.

To summarize, the design objectives of PostSharp 2.0 are:

  • To reliably support multi-vendor solutions, implying:
  • o Strict binary backward compatibility,
  • o Aspect dependency checking;
  • To implement a large vision of aspect-oriented programming (many features);
  • To improve productivity of developers working with aspects (IDE support);
  • To improve runtime performance.

Next week I will start to blog technically about new features.

Your Responsibility

I opened this post by reminding the greater responsibility coming from the position of leading aspect framework. As a closing remark, I wanted to share a part of this responsibility with you: as an early adopter of aspect-oriented technologies, your feedback will be vital. You know how communication is like: people will blog about it, say it's great (it is), and so on. I will do the same because I need to sell. But don't forget we are first engineers and not marketers.

I want you to break the product. Find flaws. Find the weak element. Where will the design break, when it will break? Compare the features to your cases. Is there a case we could solve with a small design modification? We will maybe not solve your use case in this release, but make sure the design allows us to address the case in a future release. Nobody else than you, knowledgeable of your particular business, can provide this feedback.

Thank you.

Happy PostSharping!

-gael