Is there a way to use Property Injection in Ninject 2 without using the [Inject] attribute? This creates a dependency to Ninject in the class that will be wired using it and I prefer to avoid having unneeded dependencies to my IoC container, that's why I end up using Constructor Injection more often.
I guess the same applies to Method Injection
I followed Ruben's tip and posted a small blog post on how to achieve this, but here's the quick answer:
Create a custom attribute:
public class InjectHereAttribute : Attribute
{
}
The target class will now look like this:
public class Samurai
{
[InjectHere]
public IWeapon Context { get; set; }
}
Now Ninject must be configured to use the custom attribute, this can be done by creating an implementation of IInjectionHeuristic that recognizes the custom attribute:
public class CustomInjectionHeuristic : NinjectComponent, IInjectionHeuristic, INinjectComponent, IDisposable
{
public new bool ShouldInject(MemberInfo member)
{
return member.IsDefined(
typeof(InjectHereAttribute),
true);
}
}
And finally add this behavior to the Ninject Kernel using the Components collection, it will run along the existing components, namely the default implementation of IInjectionHeuristic, which means either the default or the custom attribute can be used.
// Add custom inject heuristic
kernel.Components.Add<IInjectionHeuristic, CustomInjectionHeuristic>();
You can pass in another [attribute] type to the Kernel upon creation which can be used instead of InjectAttribute, but you'll still have to reference something centrally OOTB.
There was a similar question very recently about doing PI without attributes - there's no OOTB (as in directly on the fluent configuration interface) to put in a custom scanner but the extensibility points (you add a component that implements a Ninject interface as you build your Kernel that dictates how that aspect is to be work if looking for a given attribute isnt't what you want) are in there to determine where to inject based on Convention over Configuration - there's nothing stopping you amending the scanning to be based on just an attribute name (so it doesnt necessarily have to live in a central location).
Note that, in general, constructor injection is good for lots of reasons anyway, including this one, and keeping you code container agnostic is important (even if you're currently happy with one!)
I was able to accomplish this using a Heuristic class:
public sealed class MyInjectionHeuristic : NinjectComponent, IInjectionHeuristic
{
private static readonly IList<Type>
_propertyInjectible =
new List<Type>
{
typeof(IMyService),
};
/// <summary>
/// Returns a value indicating whether the specified member should be injected.
/// </summary>
/// <param name="member">The member in question.</param>
/// <returns><c>True</c> if the member should be injected; otherwise <c>false</c>.</returns>
public bool ShouldInject(MemberInfo member)
{
var info = member as PropertyInfo;
if( member == null || info == null )
return false;
if (info.CanWrite)
return _propertyInjectible.Contains(info.PropertyType);
if( this.Settings == null )
return false;
var propList = member.GetCustomAttributes(this.Settings.InjectAttribute, true);
return propList.Length > 0;
}
}
When creating your kernel:
var heuristics = _kernel.Components.Get<ISelector>().InjectionHeuristics;
heuristics.Add(new MyInjectionHeuristic());
Simple add additional types to the IList when you want to inject other types via properties.
Related
I am trying to create a generic controller like this:
[Route("api/[controller]")]
public class OrdersController<T> : Controller where T : IOrder
{
[HttpPost("{orderType}")]
public async Task<IActionResult> Create(
[FromBody] Order<T> order)
{
//....
}
}
I intend for the {orderType} URI segment variable to control the generic type of the controller. I'm experimenting with both a custom IControllerFactory and IControllerActivator, but nothing is working. Every time I try to send a request, I get a 404 response. The code for my custom controller factory (and activator) is never executed.
Evidently the problem is that ASP.NET Core expects valid controllers to end with the suffix "Controller", but my generic controller instead has the (reflection based) suffix "Controller`1". Thus the attribute-based routes it declares are going unnoticed.
In ASP.NET MVC, at least in its early days, the DefaultControllerFactory was responsible for discovering all the available controllers. It tested for the "Controller" suffix:
The MVC framework provides a default controller factory (aptly named DefaultControllerFactory) that will search through all the assemblies in an appdomain looking for all types that implement IController and whose name ends with "Controller."
Apparently, in ASP.NET Core, the controller factory no longer has this responsibility. As I stated earlier, my custom controller factory executes for "normal" controllers, but is never invoked for generic controllers. So there is something else, earlier in the evaluation process, which governs the discovery of controllers.
Does anyone know what "service" interface is responsible for that discovery? I don't know the customization interface or "hook" point.
And does anyone know of a way to make ASP.NET Core "dump" the names of all the controllers it discovered? It would be great to write a unit test that verifies that any custom controller discovery I expect is indeed working.
Incidentally, if there is a "hook" which allows generic controller names to be discovered, it implies that route substitutions must also be normalized:
[Route("api/[controller]")]
public class OrdersController<T> : Controller { }
Regardless of what value for T is given, the [controller] name must remain a simple base-generic name. Using the above code as an example, the [controller] value would be "Orders". It would not be "Orders`1" or "OrdersOfSomething".
Note
This problem could also be solved by explicitly declaring the closed-generic types, instead of generating them at run time:
public class VanityOrdersController : OrdersController<Vanity> { }
public class ExistingOrdersController : OrdersController<Existing> { }
The above works, but it produces URI paths that I don't like:
~/api/VanityOrders
~/api/ExistingOrders
What I had actually wanted was this:
~/api/Orders/Vanity
~/api/Orders/Existing
Another adjustment gets me the URI's I'm looking for:
[Route("api/Orders/Vanity", Name ="VanityLink")]
public class VanityOrdersController : OrdersController<Vanity> { }
[Route("api/Orders/Existing", Name = "ExistingLink")]
public class ExistingOrdersController : OrdersController<Existing> { }
However, although this appears to work, it does not really answer my question. I would like to use my generic controller directly at run-time, rather than indirectly (via manual coding) at compile-time. Fundamentally, this means I need ASP.NET Core to be able to "see" or "discover" my generic controller, despite the fact that its run-time reflection name does not end with the expected "Controller" suffix.
What happens by default
During the controller discovery process, your open generic Controller<T> class will be among the candidate types. But the default implementation of the IApplicationFeatureProvider<ControllerFeature> interface, DefaultControllerTypeProvider, will eliminate your Controller<T> because it rules out any class with open generic parameters.
Why overriding IsController() doesn't work
Replacing the default implementation of the IApplicationFeatureProvider<ControllerFeature> interface, in order to override DefaultControllerTypeProvider.IsController(), will not work. Because you don't actually want the discovery process to accept your open generic controller (Controller<T>) as a valid controller. It is not a valid controller per se, and the controller factory wouldn't know how to instantiate it anyway, because it wouldn't know what T is supposed to be.
What needs to be done
1. Generate closed controller types
Before the controller discovery process even starts, you need to generate closed generic types from your open generic controller, using reflection. Here, with two sample entity types, named Account and Contact:
Type[] entityTypes = new[] { typeof(Account), typeof(Contact) };
TypeInfo[] closedControllerTypes = entityTypes
.Select(et => typeof(Controller<>).MakeGenericType(et))
.Select(cct => cct.GetTypeInfo())
.ToArray();
We now have closed TypeInfos for Controller<Account> and Controller<Contact>.
2. Add them to an application part and register it
Application parts are usually wrapped around CLR assemblies, but we can implement a custom application part providing a collection of types generated at runtime. We simply need to have it implement the IApplicationPartTypeProvider interface. Therefore, our runtime-generated controller types will enter the controller discovery process like any other built-in type would.
The custom application part:
public class GenericControllerApplicationPart : ApplicationPart, IApplicationPartTypeProvider
{
public GenericControllerApplicationPart(IEnumerable<TypeInfo> typeInfos)
{
Types = typeInfos;
}
public override string Name => "GenericController";
public IEnumerable<TypeInfo> Types { get; }
}
Registration in MVC services (Startup.cs):
services.AddMvc()
.ConfigureApplicationPartManager(apm =>
apm.ApplicationParts.Add(new GenericControllerApplicationPart(closedControllerTypes)));
As long as your controller derives from the built-in Controller class, there is no actual need to override the IsController method of the ControllerFeatureProvider. Because your generic controller inherits the [Controller] attribute from ControllerBase, it will be accepted as a controller in the discovery process regardless of its somewhat bizarre name ("Controller`1").
3. Override the controller name in the application model
Nevertheless, "Controller`1" is not a good name for routing purposes. You want each of your closed generic controllers to have independent RouteValues. Here, we will replace the name of the controller with that of the entity type, to match what would happen with two independent "AccountController" and "ContactController" types.
The model convention attribute:
public class GenericControllerAttribute : Attribute, IControllerModelConvention
{
public void Apply(ControllerModel controller)
{
Type entityType = controller.ControllerType.GetGenericArguments()[0];
controller.ControllerName = entityType.Name;
}
}
Applied to the controller class:
[GenericController]
public class Controller<T> : Controller
{
}
Conclusion
This solution stays close to the overall ASP.NET Core architecture and, among other things, you will keep full visibility of your controllers through the API Explorer (think "Swagger").
It has been tested successfully with both conventional and attribute-based routing.
Short Answer
Implement IApplicationFeatureProvider<ControllerFeature>.
Question and Answer
Does anyone know what "service" interface is responsible for [discovering all available controllers]?
The ControllerFeatureProvider is responsible for that.
And does anyone know of a way to make ASP.NET Core "dump" the names of all the controllers it discovered?
Do that within ControllerFeatureProvider.IsController(TypeInfo typeInfo).
Example
MyControllerFeatureProvider.cs
using System;
using System.Linq;
using System.Reflection;
using Microsoft.AspNetCore.Mvc.Controllers;
namespace CustomControllerNames
{
public class MyControllerFeatureProvider : ControllerFeatureProvider
{
protected override bool IsController(TypeInfo typeInfo)
{
var isController = base.IsController(typeInfo);
if (!isController)
{
string[] validEndings = new[] { "Foobar", "Controller`1" };
isController = validEndings.Any(x =>
typeInfo.Name.EndsWith(x, StringComparison.OrdinalIgnoreCase));
}
Console.WriteLine($"{typeInfo.Name} IsController: {isController}.");
return isController;
}
}
}
Register it during startup.
public void ConfigureServices(IServiceCollection services)
{
services
.AddMvcCore()
.ConfigureApplicationPartManager(manager =>
{
manager.FeatureProviders.Add(new MyControllerFeatureProvider());
});
}
Here is some example output.
MyControllerFeatureProvider IsController: False.
OrdersFoobar IsController: True.
OrdersFoobarController`1 IsController: True.
Program IsController: False.
<>c__DisplayClass0_0 IsController: False.
<>c IsController: False.
And here is a demo on GitHub. Best of luck.
Edit - Adding Versions
.NET Version
> dnvm install "1.0.0-rc2-20221" -runtime coreclr -architecture x64 -os win -unstable
NuGet.Config
<?xml version="1.0" encoding="utf-8"?>
<configuration>
<packageSources>
<clear/>
<add key="AspNetCore"
value="https://www.myget.org/F/aspnetvnext/api/v3/index.json" />
</packageSources>
</configuration>
.NET CLI
> dotnet --info
.NET Command Line Tools (1.0.0-rc2-002429)
Product Information:
Version: 1.0.0-rc2-002429
Commit Sha: 612088cfa8
Runtime Environment:
OS Name: Windows
OS Version: 10.0.10586
OS Platform: Windows
RID: win10-x64
Restore, Build, and Run
> dotnet restore
> dotnet build
> dotnet run
Edit - Notes on RC1 vs RC2
This might not be possible is RC1, because DefaultControllerTypeProvider.IsController() is marked as internal.
Application Feature Providers examine application parts and provide features for those parts. There are built-in feature providers for the following MVC features:
Controllers
Metadata Reference
Tag Helpers
View Components
Feature providers inherit from IApplicationFeatureProvider, where T is the type of the feature. You can implement your own feature providers for any of MVC's feature types listed above. The order of feature providers in the ApplicationPartManager.FeatureProviders collection can be important, since later providers can react to actions taken by previous providers.
By default, ASP.NET Core MVC ignores generic controllers (for example, SomeController). This sample uses a controller feature provider that runs after the default provider and adds generic controller instances for a specified list of types (defined in EntityTypes.Types):
public class GenericControllerFeatureProvider : IApplicationFeatureProvider<ControllerFeature>
{
public void PopulateFeature(IEnumerable<ApplicationPart> parts, ControllerFeature feature)
{
// This is designed to run after the default ControllerTypeProvider,
// so the list of 'real' controllers has already been populated.
foreach (var entityType in EntityTypes.Types)
{
var typeName = entityType.Name + "Controller";
if (!feature.Controllers.Any(t => t.Name == typeName))
{
// There's no 'real' controller for this entity, so add the generic version.
var controllerType = typeof(GenericController<>)
.MakeGenericType(entityType.AsType()).GetTypeInfo();
feature.Controllers.Add(controllerType);
}
}
}
}
The entity types:
public static class EntityTypes
{
public static IReadOnlyList<TypeInfo> Types => new List<TypeInfo>()
{
typeof(Sprocket).GetTypeInfo(),
typeof(Widget).GetTypeInfo(),
};
public class Sprocket { }
public class Widget { }
}
The feature provider is added in Startup:
services.AddMvc()
.ConfigureApplicationPartManager(p =>
p.FeatureProviders.Add(new GenericControllerFeatureProvider()));
By default, the generic controller names used for routing would be of the form GenericController`1[Widget] instead of Widget. The following attribute is used to modify the name to correspond to the generic type used by the controller:
using Microsoft.AspNetCore.Mvc.ApplicationModels;
using System;
namespace AppPartsSample
{
// Used to set the controller name for routing purposes. Without this convention the
// names would be like 'GenericController`1[Widget]' instead of 'Widget'.
//
// Conventions can be applied as attributes or added to MvcOptions.Conventions.
[AttributeUsage(AttributeTargets.Class, AllowMultiple = false, Inherited = true)]
public class GenericControllerNameConvention : Attribute, IControllerModelConvention
{
public void Apply(ControllerModel controller)
{
if (controller.ControllerType.GetGenericTypeDefinition() !=
typeof(GenericController<>))
{
// Not a GenericController, ignore.
return;
}
var entityType = controller.ControllerType.GenericTypeArguments[0];
controller.ControllerName = entityType.Name;
}
}
}
The GenericController class:
using Microsoft.AspNetCore.Mvc;
namespace AppPartsSample
{
[GenericControllerNameConvention] // Sets the controller name based on typeof(T).Name
public class GenericController<T> : Controller
{
public IActionResult Index()
{
return Content($"Hello from a generic {typeof(T).Name} controller.");
}
}
}
Sample: Generic controller feature
To get a list of controllers in RC2, just get ApplicationPartManager from DependencyInjection and do this:
ApplicationPartManager appManager = <FROM DI>;
var controllerFeature = new ControllerFeature();
appManager.PopulateFeature(controllerFeature);
foreach(var controller in controllerFeature.Controllers)
{
...
}
I'm using Ninject in an MVC project and am trying to implement Domain Events following Udi Dahan's pattern http://www.udidahan.com/2009/06/14/domain-events-salvation/
In the extract below, the "Container" is used to resolve all the event-handlers for the particular type of event that has been raised.
My question (& apologies if I am missing something basic) is how to do this with Ninject? In other words:
How does the "Container" get set in this static class?
Once I have a Container (Kernel?) what would be the Ninject syntax to resolve all the event handlers (which I'm assuming I would have to register before-hand in a Service Module)?
I keep reading in posts that only constructor injection should be used and everything recursively get resolved from that, and that accessing the Ninject Kernel is a no-no. So any advice on how to do this will be much appreciated.
Extract from the article
public static class DomainEvents
{
[ThreadStatic] //so that each thread has its own callbacks
private static List<Delegate> actions;
public static IContainer Container { get; set; } //as before
//Registers a callback for the given domain event
public static void Register<T>(Action<T> callback) where T : IDomainEvent
{
if (actions == null)
actions = new List<Delegate>();
actions.Add(callback);
}
//Clears callbacks passed to Register on the current thread
public static void ClearCallbacks ()
{
actions = null;
}
//Raises the given domain event
public static void Raise<T>(T args) where T : IDomainEvent
{
if (Container != null)
foreach(var handler in Container.ResolveAll<Handles<T>>())
handler.Handle(args);
if (actions != null)
foreach (var action in actions)
if (action is Action<T>)
((Action<T>)action)(args);
}
}
How does the "Container" get set in this static class?
You will have to set it during application startup:
DomainEvents.Container = kernel;
what would be the Ninject syntax to resolve all the event handlers:
You can do it like this, for instance:
Container.Get<IEnumerable<Handles<T>>>())
Udi's static DomainEvents class is an implementation of the Ambient Context anti-pattern (see DI PP&P chapter 5.3). In this case I would rather use dependency injection to inject an IDomainEvents abstraction into code that needs it, instead of letting code depend on a static instance.
The problem however is that your domain objects will need a dependency on the IDomainEvents and constructor injection is (probably) not possible. The trick is to use method injection in that case.
In other words, use constructor injection to inject the IDomainEvents into command handlers or services (or what ever you call your business logic that uses the methods on your domain objects) and pass that dependency into the domain object when calling a method that needs it (method injection).
In my asp.net mvc controller`s constructor I have multiple (5) interfaces which communicate with my database in this way:
[HttpGet]
public ActionResult Create()
{
var releases = _releaseDataProvider.GetReleases();
var templates = _templateDataProvider.GetTemplates();
var createTestplanViewModel = new CreateTestplanViewModel(templates, releases);
return PartialView(createTestplanViewModel);
}
Above I use 2 different interfaces to get data from the database.
business case: To create a testplan I need to show the user the available releases + templates he can choose from.
How can I decrease the dependency/over-injection of these 2 interfaces
In the MVC project:
public class MyController : Controller
{
private readonly IQueryProcessor _queryProcessor;
public MyController(IQueryProcessor queryProcessor)
{
_queryProcessor = queryProcessor;
}
[HttpGet]
public ActionResult Create()
{
var releases = _queryProcessor.Execute(new ProvideReleaseData());
var templates = _queryProcessor.Execute(new ProvideTemplateData());
var createTestplanViewModel = AutoMapper.Mapper
.Map<CreateTestplanViewModel>(releases);
AutoMapper.Mapper.Map(templates, createTestplanViewModel);
return PartialView(createTestplanViewModel);
}
}
You can then constructor inject your current provider implementations into IQueryHandler implementations. The IQueryProcessor is just infrastructure. See this for more info: https://cuttingedge.it/blogs/steven/pivot/entry.php?id=92
Reply to comments:
It's at the site I linked to. Here's mine:
using System.Diagnostics;
using SimpleInjector;
namespace MyApp.Infrastructure
{
sealed class SimpleQueryProcessor : IQueryProcessor
{
private readonly Container _container;
public SimpleQueryProcessor(Container container)
{
_container = container;
}
[DebuggerStepThrough]
public TResult Execute<TResult>(IDefineQuery<TResult> query)
{
var handlerType = typeof(IHandleQueries<,>)
.MakeGenericType(query.GetType(), typeof(TResult));
dynamic handler = _container.GetInstance(handlerType);
return handler.Handle((dynamic)query);
}
}
}
A good general way to decouple your database would be using a unit of work. Here's a great article on from asp.net, as well as another article on MSDN.
In summary, you create a single unit where all of your database/service calls reside and it can handle the database logic. This would reduce the dependancy of your multiple interfaces into a single point, so you would only need to inject 1 class into your controller.
A quote from the MSDN article:
According to Martin Fowler, the Unit of Work pattern "maintains a list
of objects affected by a business transaction and coordinates the
writing out of changes and the resolution of concurrency problems."
EDIT
It seems to me you basically have these options to reduce constructor dependency count here:
Split the controller
Add layer in front of the two interfaces
Switch to property injection
Service locator
#3 and #4 are included for good measure, but they obviously don't actually decrease the dependency count, they only hide them from the constructor. They also have several disadvantages, and I consider service locator especially evil most of the time.
For #1, if you feel your constructor is actually doing two+ jobs, and there is a clean separation where you could split, I would do so. I assume from your responses that you have already considered this, however, and don't want to do this.
That leaves #2 - adding another layer. In this case that would be introducing a factory interface for that particular view model. Naively, I'll name this ICreateTestplanViewModelFactory, but you can name it something more sensical for your app if you wish. A single method on it would construct a CreateTestplanViewModel.
This makes the fact that the data for this view is coming from 2 sources merely an implementation detail. You would wire up an implementation which takes IReleaseDataProvider and ITemplateDataProvider as constructor dependencies.
This is along the lines of what I was suggesting:
public interface IProvideTestPlanSetupModel
{
CreateTestplanViewModel GetModel();
}
public class TestPlanSetupProvider : IProvideTestPlanSetupModel
{
private readonly IReleaseDataProvider _releaseDataProvider;
private readonly ITemplateDataProvider _templateDataProvider;
public TestPlanSetupProvider(IReleaseDataProvider releaseDataProvider, ITemplateDataProvider templateDataProvider)
{
_releaseDataProvider = releaseDataProvider;
_templateDataProvider = templateDataProvider;
}
public CreateTestplanViewModel GetModel()
{
var releases = _releaseDataProvider.GetReleases();
var templates = _templateDataProvider.GetTemplates();
return new CreateTestplanViewModel(releases, templates);
}
}
public class TestPlanController : Controller
{
private readonly IProvideTestPlanSetupModel _testPlanSetupProvider;
public TestPlanController(IProvideTestPlanSetupModel testPlanSetupProvider)
{
_testPlanSetupProvider = testPlanSetupProvider;
}
[HttpGet]
public ActionResult Create()
{
var createTestplanViewModel = _testPlanSetupProvider.GetModel();
return PartialView(createTestplanViewModel);
}
}
If you don't like constructing a view model anywhere outside the controller, the interface could provide an intermediate object with the same properties that you would copy to the view model. But that is silly, as this combination of data is only relevant for that particular view, which is precisely what the view model is supposed to represent.
On a side note, it seems you are running into pretty common annoyances doing read/write through the same model. Since these issues bother you so, you might investigate CQRS, which perhaps would make you feel less dirty about talking to the database directly for these types of queries and would help you get around the layering labyrinth we all enjoy so much. It seems promising, though I have not yet had the pleasure of test driving it in a production application.
I am writing a library that will provide a collection of public types to its consumers.
I want to make types from this library dependency injection friendly. This means that every class needs to have a constructor through which it is possible to specify every single dependency of the object being initialized. I also want the library to adhere to the convention over configuration principle. This means that if a consumer wants the default behavior, he may use a parameterless constructor and the object will somehow construct the dependencies for itself.
In example (C#):
public class Samurai {
private readonly IWeapon _weapon;
// consumers will use this constructor most of the time
public Samurai() {
_weapon = ??? // get an instance of the default weapon somehow
}
// consumers will use this constructor if they want to explicitly
// configure dependencies for this instance
public Samurai(IWeapon weapon) {
_weapon = weapon;
}
}
My first solution would be to use the service locator pattern.
The code would look like this:
...
public Samurai() {
_weapon = ServiceLocator.Instance.Get<IWeapon>();
}
...
I have a problem with this, though. Service locator has been flagged as an anti-pattern (link) and I completely agree with these arguments. On the other hand, Martin Fowler advocates use of the service locator pattern exactly in this situation (library projects) (link). I want to be careful and eliminate the possible necessity to rewrite the library after it shows up that service locator really was a bad idea.
So in conclusion - do you think that service locator is fine in this scenario? Should I solve my problem in a completely different way? Any thought is welcome...
If you want to make life easier for users who are not using a DI container, you can provide default instances via a dedicated Defaults class which has methods like this:
public virtual Samurai CreateDefaultSamurai()
{
return new Samurai(CreateDefaultWeapon());
}
public virtual IWeapon CreateDefaultWeapon()
{
return new Shuriken();
}
This way you don't need to pollute the classes themselves with default constructors, and your users aren't at risk of using those default constructors unintentionally.
There is an alternative, that is injecting a specific provider, let's say a WeaponProvider in your case into your class so it can do the lookup for you:
public interface IWeaponProvider
{
IWeapon GetWeapon();
}
public class Samurai
{
private readonly IWeapon _weapon;
public Samurai(IWeaponProvider provider)
{
_weapon = provider.GetWeapon();
}
}
Now you can provide a local default provider for a weapon:
public class DefaultWeaponProvider : IWeaponProvider
{
public IWeapon GetWeapon()
{
return new Sword();
}
}
And since this is a local default (as opposed to one from a different assembly, so it's not a "bastard injection"), you can use it as part of your Samurai class as well:
public class Samurai
{
private readonly IWeapon _weapon;
public Samurai() : this(new DefaultWeaponProvider())
{
}
public Samurai(IWeaponProvider provider)
{
_weapon = provider.GetWeapon();
}
}
I have used the following approach in my C# project. The goal was to achieve dependency injection (for unit / mock testing) whilst not crippling the implementation of the code for a "normal use case" (i.e. having a large amount of new()'s that are cascaded through the execution flow).
public sealed class QueueProcessor : IQueueProcessor
{
private IVbfInventory vbfInventory;
private IVbfRetryList vbfRetryList;
public QueueProcessor(IVbfInventory vbfInventory = null, IVbfRetryList vbfRetryList = null)
{
this.vbfInventory = vbfInventory ?? new VbfInventory();
this.vbfRetryList = vbfRetryList ?? new VbfRetryList();
}
}
This allows DI but also means any consumer doesn't have to worry about what the "default instance flow" should be.
Let's say we have class:
public class WithDependencies
{
public WithDependencies(IAmDependencyOne first, IAmDependencyTwo second)
// ...
}
Now the question. How do you create objects of WithDependencies class in an application?
I know there are many ways.
new WithDependencies(new DependencyOne(), new DependencyTwo());
new WithDependencies(IoC.Resolve(IDependencyOne), IoC.Resolve(IDependencyTwo());
// register IDependencyOne, IDependencyTwo implementations at app start
IoC.Resolve(WithDependencies);
// register IDependencyOne, IDependencyTwo implementations at app start
// isolate ourselves from concrete IoC Container
MyCustomWithDependenciesFactory.Create();
and so on...
What do you think is the way to do it?
Edit:
Because I don't get answers or I don't understand them I'll try to ask again. Let's say that on some event (button, timer, whatever) I need new object WithDependencies(). How do I create it? Assume IoC container is already configured.
It depends on the context, so it's impossible to provide a single answer. Conceptually you'd be doing something like this from the Composition Root:
var wd = new WithDependencies(new DependencyOne(), new DependencyTwo());
However, even in the absence of a DI Container, the above code isn't always unambiguously the correct answer. In some cases, you might want to share the same dependency among several consumers, like this:
var dep1 = new DependencyOne();
var wd = new WithDependencies(dep1, new DependencyTwo());
var another = AnotherWithDependencies(dep1, new DependencyThree());
In other cases, you might not want to share dependencies, in which case the first option is more correct.
This is just a small glimpse of an entire dimension of DI concerned with Lifetime Management. Many DI Containers can take care of that for you, which is one excellent argument to prefer a DI Container over Poor Man's DI.
Once you start using a DI Container, you should follow the Register Resolve Release pattern when resolving types, letting Auto-wiring take care of the actual composition:
var wd = container.Resolve<WithDependencies>();
The above example assumes that the container is already correctly configured.
If you need to create a dependency which has its own dependencies, you can either A) do it yourself, or B) ask something else to do it for you. Option A negates the benefits of dependency injection (decoupling, etc.), so I would say option B is a better starting point. Now, we have chosen to use the factory pattern, no matter whether it takes the form of a service locator (i.e. IoC.Resolve), a static factory, or an instance factory. The point is that we have delegated that responsibility to an external authority.
There are a number of trade-offs required for static accessors. (I went over them in another answer, so I won't repeat them here.) In order to avoid introducing a dependency on the infrastructure or the container, a solid option is to accept a factory for creating WithDependencies when we need an instance somewhere else:
public class NeedsWithDependencies
{
private readonly IWithDependenciesFactory _withDependenciesFactory;
public NeedsWithDependencies(IWithDependenciesFactory withDependenciesFactory)
{
_withDependenciesFactory = withDependenciesFactory;
}
public void Foo()
{
var withDependencies = _withDependenciesFactory.Create();
...Use the instance...
}
}
Next, we can create a container-specific implementation of the factory:
public class WithDependenciesFactory : IWithDependenciesFactory
{
private readonly IContainer _container;
public WithDependenciesFactory(IContainer container)
{
_container = container
}
public WithDependencies Create()
{
return _container.Resolve<WithDependencies>();
}
}
Now NeedsWithDependencies is completely isolated from any knowledge of how WithDependencies gets created; it also exposes all its dependencies in its constructor, instead of hiding dependencies on static accessors, making it easy to reuse and test.
Defining all those factories can get a little cumbersome, though. I like Autofac's factory relationship type, which will detect parameters of the form Func<TDependency> and automatically inject a function which serves the same purpose as the hand-coded factory above:
public class NeedsWithDependencies
{
private readonly Func<WithDependencies> _withDependenciesFactory;
public NeedsWithDependencies(Func<WithDependencies> withDependenciesFactory)
{
_withDependenciesFactory = withDependenciesFactory;
}
public void Foo()
{
var withDependencies = _withDependenciesFactory();
...Use the instance...
}
}
It also works great with runtime parameters:
public class NeedsWithDependencies
{
private readonly Func<int, WithDependencies> _withDependenciesFactory;
public NeedsWithDependencies(Func<int, WithDependencies> withDependenciesFactory)
{
_withDependenciesFactory = withDependenciesFactory;
}
public void Foo(int x)
{
var withDependencies = _withDependenciesFactory(x);
...Use the instance...
}
}
Sometimes I try to get rid of factories or at least not depend directly on them, so Dependency Injection (without factories) is useful of course.
Therefore I use Google Juice, cause its a small little framework using Java Annotations and you can quickly change your injections / dependencies. Just take a look at it:
http://code.google.com/p/google-guice/