Is it possible to configure Unity to either detect a circular reference or to intercept the type resolver to display some debugging information?
Example
Here are a couple of interfaces and classes which are dependent upon each other
public interface IThing1 { }
public class Thing1 : IThing1
{
private IThing2 _thing2;
public Thing1(IThing2 thing2)
{
_thing2 = thing2;
}
}
public interface IThing2 { }
public class Thing2 : IThing2
{
private IThing1 _thing1;
public Thing2(IThing1 thing1)
{
_thing1 = thing1;
}
}
Castle Windsor
If these two types are configured in Castle Windsor it will throw an exception and provide some debug information to find the circular reference:
Castle.MicroKernel.CircularDependencyException: Dependency cycle has been detected when trying to resolve component 'CircularIoC.Thing1'.
The resolution tree that resulted in the cycle is the following:
Component 'CircularIoC.Thing1' resolved as dependency of
component 'CircularIoC.Thing2' resolved as dependency of
component 'CircularIoC.Thing1' which is the root component being resolved.
Unity
If Unity is configured to resolve these types like so
private static void ResolveWithUnity()
{
var container = new UnityContainer();
container.RegisterType<IThing1, Thing1>();
container.RegisterType<IThing2, Thing2>();
var thing = container.Resolve<IThing1>();
container.Dispose();
}
The call to container.Resolve<> will cause a StackOverflowException.
This is the documented behaviour but it would be nice to have some more useful information. Is there any customisation that will provide more information about the circular reference?
Alternatively is there any way to hook in to the type resolver process to emit some debugging information? I am thinking of decorating the main type resolver to output the name of the type being resolved. This will provide some feedback and a pointer to dependency that is causing the circular reference.
While I know that changing to a different IoC would solve the problem, this is not unfortunately an option.
Unity sadly doesn't support this (incredibly important) feature.
If you are willing to put your back into it, you can implement a smart decorator using some elbow grease.
What you will need is to override all Registration methods and build and update a data structure for the dependencies (dependency graph).
Then write a method that preforms DFS to detect a circular dependency, you can either use it as a finalizer for the registration process, that will detect pre-resolving if circular dependencies are possible, or use it per resolve for the specific type requested.
As you can see, it's a lot of work...
Another option, is just to wrap up with a decorator the resolve methods and catch the StackOverflowException, analyze it to make sure it resulted from the resolving process, and build a proper circular dependency exception.
Related
This involves autofac and c#. I have an interface derived from a parent interface:
public interface IJ4JLogger<out TCalling>
{
}
public interface IJ4JSmsLogger<out TCalling> : IJ4JLogger<TCalling>
{
}
Certain classes depend on being supplied an instance of the parent interface during construction:
public FileHistoryConfiguration( IJ4JLogger<FileHistoryConfiguration> histLogger, IJ4JLogger<FileHistoryService> svcLogger )
{
}
But if I register the type like this with autofac:
builder.RegisterGeneric( typeof(J4JSmsLogger<>) )
.As(typeof(IJ4JSmsLogger<>))
.SingleInstance();
where J4JSmsLogger<> is a class implementing IJ4JSmsLogger<>, then this call fails with an error that it can't find anything registered to provide an IJ4JLogger<> interface:
_fhConfig = _svcProvider.GetRequiredService<IFileHistoryConfiguration>();
I can work around the problem by changing the As<> clause in the registration of J4JSmsLogger<> to treat it as a IJ4JLogger<> instance, and then cast the result of resolving that interface to IJ4JSmsLogger<> whenever I need the extra capabilities of the child interface.
But I don't understand why I have to do that. Is there an additional step I need to take during registration of the types with autofac so that objects implementing the child interface will satisfy a need for the parent interface?
Cleaner Workaround
Reading more about autofac I learned something new: you can define as many As<>() clauses (including AsSelf()) as you want. So changing my autofac configuration to:
builder.RegisterGeneric( typeof(J4JSmsLogger<>) )
.As(typeof(IJ4JSmsLogger<>))
.As(typeof(IJ4JLogger<>))
.SingleInstance();
provides a cleaner solution than constantly casting resolved instances.
I'm not going to submit it as an answer, though, because I am curious why autofac doesn't do this kind of downcasting automatically, and whether any other DI frameworks do.
Autofac won't cast to base types for you like that. It generally assumes wiring is exact. You could run into some real problems if it didn't, like if someone has a constructor like...
public class BadTimes
{
public BadTimes(object input) { }
}
Which object does it put in there? Everything casts down to object.
However, you could always register it as both types and call it a day:
builder.RegisterGeneric(typeof(J4JSmsLogger<>))
.As(typeof(IJ4JSmsLogger<>))
.As(typeof(IJ4JLogger<>))
.SingleInstance();
We're refactoring an older system to use DI. Sadly, some of the "core" components that are used all over everywhere have injection unfriendly constructors (descriptions, for example), so we have to use ServiceLocator to create them. Refactoring them is very impractical at this time.
We're trying to create the unfriendly classes by injecting ILifetimeScope into the appropriate place, but are getting the following exception:
No constructors on type 'Autofac.Core.Registration.ScopeRestrictedRegistry' can be found with the constructor finder 'Autofac.Core.Activators.Reflection.DefaultConstructorFinder'.
If I cheat and use the "Update" method on the ContainerBuilder and then register the container as the LifetimeScope, the resolution works successfully, however, given that Update is obsolete, it's not something I want to do.
Can anyone help?
Edit: I'm not doing anything special. Build up is standard:
builder.RegisterAssemblyTypes(assembly).AsImplementedInterfaces();
this.Container = builder.Build();
builder = new ContainerBuilder();
builder.RegisterInstance(this.Container);
builder.RegisterInstance(this.Container).As<ILifetimeScope>();
builder.Update(this.Container);
Without these lines
builder = new ContainerBuilder();
builder.RegisterInstance(this.Container);
builder.RegisterInstance(this.Container).As<ILifetimeScope>();
builder.Update(this.Container);
any class with an ILifetimeScope dependency fails with the error above.
public class MyClass : IMyClass
{
public MyClass(ILifetimeScope scope)
{
...
}
}
I'm actually thinking that this is a bug in the Autofac Framework, so I'm hoping that someone from the team will be able to tell me more.
ILifetimeScope is supposed to automatically be available.
**I'm using PostSharp Express... not sure that would make a difference in this instance though.
I've got an OnMethodBoundary->OnEntry aspect that successfully multicasts at the assembly level to class members in my own code, but when I attempt to apply it to mscorlib System.IO.StreamReader members, no dice. Based on the searching I've done on the PostSharp web site, here on SO, and on Google, I can't tell what the correct way to go about this is with the current version of PostSharp. Does anyone know? Hopefully I'm just missing something simple :\
Here's the aspect followed by multicast attribute I'm using:
namespace Test.Aspects {
[AttributeUsage(AttributeTargets.Assembly)]
[MulticastAttributeUsage(MulticastTargets.Method, AllowMultiple = false)]
[Serializable]
public class PatchStreamReaderAttribute : OnMethodBoundaryAspect {
public override void OnEntry(MethodExecutionArgs args) {
System.Threading.Thread.Sleep(1000);
}
}
}
[assembly: PatchStreamReader(AttributeTargetMembers = "ReadLine", AttributeTargetAssemblies = "mscorlib", AttributeTargetTypes = "System.IO.StreamReader")]
Usually, when you apply an aspect in a given assembly, PostSharp will modify that assembly during its build process. This, of course, cannot happen for mscorlib or, in fact, for any 3-rd party library you reference but do not build from source code.
This is why PostSharp uses different approach when applying aspects to the referenced assemblies using AttributeTargetAssemblies. Instead of modifying the target 3-rd party assembly, PostSharp will modify the calls from your assembly to the target assembly.
This, of course, gives you less options of where you can inject your code. For example, PostSharp can detect the call to the library's method and inject the aspect around that call. But you cannot inject the aspect around the static or instance constructor of the type from the library.
You also need to pay attention to the AttributeTargetTypes property when applying the aspect. For example, you want to apply the aspect on the calls to the StreamReader.ReadLine() method. This virtual ReadLine() method is originally declared on the TextReader class and StreamReader overrides the method. If you look at the IL, then the method call looks like this:
callvirt instance string [mscorlib]System.IO.TextReader::ReadLine()
This means you need to set AttributeTargetTypes property to "System.IO.TextReader" to apply the aspect to the ReadLine() method.
I have a WPF application based on MVVM with Caliburn.Micro and Ninject. I have a root viewmodel called ShellViewModel. It has a couple of dependencies (injected via constructor) which are configured in Caliburn's Bootstrapper. So far so good.
Somewhere down the line, there is a MenuViewModel with a couple of buttons, that in turn open other viewmodels with their own dependencies. These viewmodels are not created during creation of the root object, but I still want to inject dependencies into them from my IoC container.
I've read this question on service locator vs dependency injection and I understand the points being made.
I'm under the impression however that my MenuViewModel needs to be able to access my IoC container in order the properly inject the viewmodels that are being made dynamically..which is something I'm trying to avoid. Is there another way?
Yes, I believe you can do something a bit better.
Consider that if there was no on-demand requirement then obviously you could make those viewmodels be dependencies of MenuViewModel and so on up the chain until you get to the root of the object graph (the ShellViewModel) and the container would wire everything up.
You can put a "firewall" in the object graph by substituting something that can construct the dependencies of MenuViewModel for the dependencies themselves. The container is the obvious choice for this job, and IMHO from a practical standpoint this is a good enough solution even if it's not as pure.
But you can also substitute a special-purpose factory instead of the container; this factory would take a dependency on the container and provide read-only properties for the real dependencies of MenuViewModel. Accessing the properties would result in having the container resolve the objects and returning them (accessor methods would also work instead of properties; what's more appropriate is another discussion entirely, so just use whatever you think is better).
It may look like that you haven't really changed the status quo, but the situation is not the same it would be if MenuViewModel took a direct dependency on the container. In that case you would have no idea what the real dependencies of MenuViewModel are by looking at its public interface, while now you would see that there's a dependency on something like
interface IMenuViewModelDependencyFactory
{
public RealDependencyA { get; }
public RealDependencyB { get; }
}
which is much more informative. And if you look at the public interface of the concrete MenuViewModelDependencyFactory things are also much better:
class MenuViewModelDependencyFactory : IMenuViewModelDependencyFactory
{
private Container container;
public MenuViewModelDependencyFactory(Container container) { ... }
public RealDependencyA { get { ... } }
public RealDependencyB { get { ... } }
}
There should be no confusion over what MenuViewModelDependencyFactory intends to do with the container here because it's so very highly specialized.
I am working in a project that has two main parts: a class library assembly and the main application. Both are using Castle Windsor for IoC and both manually setup their list of of components in code (to aid refactoring and prevent the need for a config file). Currently the main application has code like this:
public static void Main()
{
// Perform library IoC setup
LibraryComponent.Init();
// Perform application IoC setup
IoC.Register<IXyz, Abc>("abc");
// etc, etc, ...
// Start the application code ...
}
However the call to initialise the library doesn't seem like a good solution. What is the best way to setup a class library that uses an IoC container to decouple its internal components?
Edit:
Lusid proposed using a static method on each public component in the library that would in turn make the call to initialise. One possible way to make this a bit nicer would be to use something like PostSharp to do this in an aspect-oriented way. However I was hoping for something a bit more elegant ;-)
Lusid also proposed using the AppDomain.AssemblyLoad event to perform custom steps at load time, however I am really after a way to avoid the client assembly from requiring any setup code.
Thanks!
I'm not sure if I'm understanding exactly the problem you are trying to solve, but my first guess is that you are looking for a way to decouple the need to call the Init method from your main application.
One method I've used in the past is a static constructor on a static class in the class library:
static public class LibraryComponent {
static LibraryComponent() {
Init();
}
}
If you have multiple class libraries, and would like a quick and dirty way of evaluating all of them as they are loaded, here's a (kinda hairy) way:
[STAThread]
static void Main()
{
AppDomain.CurrentDomain.AssemblyLoad += new AssemblyLoadEventHandler(CurrentDomain_AssemblyLoad);
}
static void CurrentDomain_AssemblyLoad(object sender, AssemblyLoadEventArgs args)
{
IEnumerable<Type> types = args.LoadedAssembly.GetTypes()
.Where(t => typeof(IMyModuleInterface).IsAssignableFrom(t));
foreach (Type t in types)
{
doSomethingWithType(t);
}
}
The Where clause could be anything you want, of course. The code above would find any class deriving from IMyModuleInterface in each assembly that gets loaded into the current AppDomain, and then I can do something with it, whether it be registering dependencies, maintaining an internal list, whatever.
Might not be exactly what you are looking for, but hopefully it helps in some way.
You could have a registration module. Basically LibraryComponent.Init() function takes an IRegistrar to wire everything up.
The IRegistrar could basically have a function Register(Type interface, Type implementation). The implimentor would map that function back to their IOC container.
The downside is that you can't rely on anything specific to the container your using.
Castle Windsor actually has a concept called facilities that are basically just ways of wrapping standardised pieces of configuration. In this model, you would simply add the two facilities to the container and they would do the work.
Of course, this wouldn't really be better than calling a library routine to do the work unless you configured the facilities in a configuration file (consider binsor). If you are really allergic to configuration files, your current solution is probably the best.