I've read a lot on the subject of IoC containers and particularly Mark Seemann's blog posts, where he emphasises the importance of convention over configuration. I understand and agree with his point, but I wonder how best to mark out a class for an irregular lifetime?
For example, most of my services are registered with a Transient lifetime - but I want a particular service to be registered as a Singleton because it does some useful caching.
I have thought about using some custom attributes, but I have read arguments against this (because it puts composition logic in the class).
I have had this same debate with, for example, dependencies on configuration primitives. Ultimately I ended up using a parameter attribute because it worked in my case, but I feel that perhaps I didn't see the 'hidden dangers'.
It is common for an IoC container to allow you to specify the lifestyle of an object. SimpleInjector, for example, provides Transient and Singleton out of the box (and also allows the user to create a custom lifestyle class). With SimpleInjector it's as simple as:
container.Register<ISvc, Impl>(Lifestyle.Singleton);
or
container.Register<ISvc, Impl>(Lifestyle.Transient);
No need for polluting your classes with composition logic.
As for configuration primitives, a common practice is to put those behind an interface. I put my connection string behind an IDatabaseConfiguration interface, for example, and its implementation reads the value from web.config. I then inject that interface into my data services.
EDIT: Disregard above. Keeping it in so that comments make sense.
The following doesn't actually answer the question, but it may solve the problem that prompted the question. You said that the class does some useful caching, which clues me in on that it is taking on an additional responsibility. I would therefore recommend that you do not attempt to register that single implementation as a singleton, but instead create a decorator class around that implementation.
container.RegisterDecorator(
typeof(ICommandHandler<>),
typeof(CommandHandlerCacheDecorator<>),
Lifestyle.Transient,
x => { return /*some logic that looks for and
finds that one class you want
to decorate*/
});
The decorator needs to be a transient, because it will have a transient reference to the command it decorates. However, it could then access a separate singleton class (via dependency injection of course) or cache object that handles the actual caching.
EDIT: Examples to single out the one class to decorate. There are plenty of options depending on just how specific you want to be. Again this is with SimpleInjector, but I'm sure other containers have their own analogues.
//decorate a specific class
x => { return x.ImplementationType.FullName == "My.Commands.Web.SomeName"; }
//decorate all classes that share a certain namespace
x => { return x.ImplementationType.Namespace.EndsWith("Commands.Web"); }
//decorate all classes that implement the same interface
x => { return x.ImplementationType.IsInstanceOfType(typeof(ICouldCache)); }
Related
I am new to dependency injection pattern. I love the idea, but struggle to apply it to my case. I have a singleton object, let’s call it X, which I need often in many parts of my program, in many different classes, sometimes deep in the call stack. Usually I would implement this as a globally available singleton. How is this implemented within the DI pattern, specifically with .NET Core DI container? I understand I need to register X with the DI container as a singleton, but how then I get access to it? DI will instantiate classes with constructors which will take reference to X, that’s great – but I need X deep within the call hierarchy, within my own objects which .NET Core or DI container know nothing about, in objects that were created using new rather than instantiated by the DI container.
I guess my question is – how does global singleton pattern aligns/implemented by/replaced by/avoided with the DI pattern?
Well, "new is glue" (Link). That means if you have new'ed an instance, it is glued to your implementation. You cannot easily exchange it with a different implementation, for example a mock for testing. Like gluing together Lego bricks.
I you want to use proper dependency injection (using a container/framework or not) you need to structure your program in a way that you don't glue your components together, but instead inject them.
Every class is basically at hierarchy level 1 then. You need an instance of your logger? You inject it. You need an instance of a class that needs a logger? You inject it. You want to test your logging mechanism? Easy, you just inject something that conforms to your logger interface that logs into a list and the at the end of your test you can check your list and see if all the required logs are there. That is something you can automate (in contrast to using your normal logging mechanism and checking the logfiles by hand).
That means in the end, you don't really have a hierarchy, because every class you have just gets their dependencies injected and it will be the container/framework or your controlling code that determines what that means for the order of instantiation of objects.
As far as design patterns go, allow me an observation: even now, you don't need a singleton. Right now in your program, it would work if you had a plain global variable. But I guess you read that global variables are "bad". And design patterns are "good". And since you need a global variable and singleton delivers a global variable, why use the "bad", when you can use the "good" right? Well, the problem is, even with a singleton, the global variable is bad. It's a drawback of the pattern, a toad you have to swallow for the singleton logic to work. In your case, you don't need the singleton logic, but you like the taste of toads. So you created a singleton. Don't do that with design patterns. Read them very carefully and make sure you use them for the intended purpose, not because you like their side-effects or because it feels good to use a design pattern.
Just an idea and maybe I need your thought:
public static class DependencyResolver
{
public static Func<IServiceProvider> GetServiceProvider;
}
Then in Startup:
public void Configure(IApplicationBuilder app, IServiceProvider serviceProvider)
{
DependencyResolver.GetServiceProvider = () => { return serviceProvider; };
}
And now in any deed class:
DependencyResolver.GetServiceProvider().GetService<IService>();
Here's a simplified example of how this would work without a singleton.
This example assumes that your project is built in the following way:
the entry point is main
main creates an instance of class GuiCreator, then calls the method createAndRunGUI()
everything else is handled by that method
So your simplified code looks like this:
// main
// ... (boilerplate)
container = new Container();
gui = new GuiCreator(container.getDatabase(), container.getLogger(), container.getOtherDependency());
gui.createAndRunGUI();
// ... (boilerplate)
// GuiCreator
public class GuiCreator {
private IDatabase db;
private ILogger log;
private IOtherDependency other;
public GuiCreator(IDatabase newdb, ILogger newlog, IOtherDependency newother) {
db = newdb;
log = newlog;
other = newother;
}
public void createAndRunGUI() {
// do stuff
}
}
The Container class is where you actually define which implementations will be used, while the GuiCreator contructor takes interfaces as arguments. Now let's say the implementation of ILogger you choose has itself a dependency, defined by an interface its contructor takes as argument. The Container knows this and resolves it accordingly by instantiating the Logger as new LoggerImplementation(getLoggerDependency());. This goes on for the entire dependency chain.
So in essence:
All classes keep instances of interfaces they depend upon as members.
These members are set in the respective constructor.
The entire dependency chain is thus resolved when the first object is instantiated. Note that there might/should be some lazy loading involved here.
The only places where the container's methods are accessed to create instances are in main and inside the container itself:
Any class used in main receives its dependencies from main's container instance.
Any class not used in main, but rather used only as a dependency, is instantiated by the container and receives its dependencies from within there.
Any class used neither in main nor indirectly as a dependency somewhere below the classes used in main will obviously never be instantiated.
Thus, no class actually needs a reference to the container. In fact, no class needs to know there even is a container in your project. All they know is which interfaces they personally need.
The Container can either be provided by some third party library/framework or you can code it yourself. Typically, it will use some configuration file to determine which implementations are actually supposed to be used for the various interfaces. Third party containers will usually perform some sort of code analysis supported by annotations to "autowire" implementations, so if you go with a ready-made tool, make sure you read up on how that part works because it will generally make your life easier down the road.
Please be gentle, I'm a newb to this IoC/MVC thing but I am trying. I understand the value of DI for testing purposes and how IoC resolves dependencies at run-time and have been through several examples that make sense for your standard CRUD operations...
I'm starting a new project and cannot come up with a clean way to accomplish user permissions. My website is mostly secured with any pages with functionality (except signup, FAQ, about us, etc) behind a login. I have a custom identity that has several extra properties which control access to data... So....
Using Ninject, I've bound a concrete type* to a method (Bind<MyIdentity>().ToMethod(c => MyIdentity.GetIdentity()); so that when I add MyIdentity to a constructor, it is injected based on the results of the method call.
That all works well. Is it appropriate to (from the GetIdentity() method) directly query the request cookies object (via FormsAuthentication)? In testing the controllers, I can pass in an identity, but the GetIdentity() method will be essentially untestable...
Also, in the GetIdentity() method, I will query the database. Should I manually create a concrete instance of a repository?
Or is there a better way all together?
I think you are reasonably on the right track, since you abstracted away database communication and ASP.NET dependencies from your unit tests. Don't worry that you can't test everything in your tests. There will always be lines of code in your application that are untestable. The GetIdentity is a good example. Somewhere in your application you need to communicate with framework specific API and this code can not be covered by your unit tests.
There might still be room for improvement though. While an untested GetIdentity isn't a problem, the fact that it is actually callable by the application. It just hangs there, waiting for someone to accidentally call it. So why not abstract the creation of identities. For instance, create an abstract factory that knows how to get the right identity for the current context. You can inject this factory, instead of injecting the identity itself. This allows you to have an implementation defined near the application's composition root and outside reach of the rest of the application. Besides that, the code communicates more clearly what is happening. Nobody has to ask "which identity do I actually get?", because it will be clear by the method on the factory they call.
Here's an example:
public interface IIdentityProvider
{
// Bit verbose, but veeeery clear,
// but pick another name if you like,
MyIdentity GetIdentityForCurrentUser();
}
In your composition root you can have an implementation of this:
private sealed class AspNetIdentityProvider : IIdentityProvider
{
public MyIdentity GetIdentityForCurrentUser()
{
// here the code of the MyIdentity.GetIdentity() method.
}
}
As a trick I sometimes have my test objects implement both the factory and product, just for convenience during unit tesing. For instance:
private sealed class FakeMyIdentity
: FakeMyIdentity, IIdentityProvider
{
public MyIdentity GetIdentityForCurrentUser()
{
// just returning itself.
return this;
}
}
This way you can just inject a FakeMyIdentity in a constructor that expects an IIdentityProvider. I found out that this doesn’t sacrifice readability of the tests (which is important).
Of course you want to have as little code as possible in the AspNetIdentityProvider, because you can't test it (automatically). Also make sure that your MyIdentity class doesn't have any dependency on any framework specific parts. If so you need to abstract that as well.
I hope this makes sense.
There are two things I'd kinda do differently here...
I'd use a custom IPrincipal object with all the properties required for your authentication needs. Then I'd use that in conjunction with custom cookie creation and the AuthenticateRequest event to avoid database calls on every request.
If my IPrincipal / Identity was required inside another class, I'd pass it as a method parameter rather than have it as a dependency on the class it's self.
When going down this route I use custom model binders so they are then parameters to my actions rather than magically appearing inside my action methods.
NOTE: This is just the way I've been doing things, so take with a grain of salt.
Sorry, this probably throws up more questions than answers. Feel free to ask more questions about my approach.
I understand the concept behind DI, but I'm just learning what different IoC containers can do. It seems that most people advocate using IoC containers to wire up stateless services, but what about using them for stateful objects like entities?
Whether it's right or wrong, I normally stuff my entities with behavior, even if that behavior requires an outside class. Example:
public class Order : IOrder
{
private string _ShipAddress;
private IShipQuoter _ShipQuoter;
public Order(IOrderData OrderData, IShipQuoter ShipQuoter)
{
// OrderData comes from a repository and has the data needed
// to construct order
_ShipAddress = OrderData.ShipAddress; // etc.
_ShipQuoter = ShipQuoter;
}
private decimal GetShippingRate()
{
return _ShipQuoter.GetRate(this);
}
}
As you can see, the dependencies are Constructor Injected. Now for a couple of questions.
Is it considered bad practice to have your entities depend on outside classes such as the ShipQuoter? Eliminating these dependencies seems to lead me towards an anemic domain, if I understand the definition correctly.
Is it bad practice to use an IoC container to resolve these dependencies and construct an entity when needed? Is it possible to do this?
Thanks for any insight.
The first question is the most difficult to answer. Is it bad practice to have Entities depend on outside classes? It's certainly not the most common thing to do.
If, for example, you inject a Repository into your Entities you effectively have an implementation of the Active Record pattern. Some people like this pattern for the convenience it provides, while others (like me) consider it a code smell or anti-pattern because it violates the Single Responsibility Principle (SRP).
You could argue that injecting other dependencies into Entities would pull you in the same direction (away from SRP). On the other hand you are certainly correct that if you don't do this, the pull is towards an Anemic Domain Model.
I struggled with all of this for a long time until I came across Greg Young's (abandonded) paper on DDDD where he explains why the stereotypical n-tier/n-layer architecture will always be CRUDy (and thus rather anemic).
Moving our focus to modeling Domain objects as Commands and Events instead of Nouns seems to enable us to build a proper object-oriented domain model.
The second question is easier to answer. You can always use an Abstract Factory to create instances at run-time. With Castle Windsor you can even use the Typed Factory Facility, relieving you of the burden of implementing the factories manually.
I know this is an old post but wanted to add. The domain entity should not persist itself even if you pass in an abstracted repository in ctor. The reason I am suggestion this is not merely that it violates SRP, it also contrary to DDD's aggregation. Let me explain, DDD is suited for complex apps with inherently deep graphs, therefore, we use aggregate or composite roots to persist changes to the underlying "children", so when we inject persistence into the individual children we violate the relationship children have to the composite or aggregate root that should be "in charge" of the life cycle or aggregation. Of course the composite root or aggregate does not persist it's own graph either. Another is with injecting dependencies of DDD objects is that an injected domain object effectively has no state until some other event takes place to hydrate its state. ANy consumer of the code will be forced to init or setup the domain object first before they can invoke business behavior which violates encapsulation.
As I understand IoC-container is helpful in creation of application-level objects like services and factories. But domain-level objects should be created manually.
Spring's manual tells us: "Typically one does not configure fine-grained domain objects in the container, because it is usually the responsibility of DAOs and business logic to create/load domain objects."
Well. But what if my domain "fine-grained" object depends on some application-level object.
For example I have an UserViewer(User user, UserConstants constants) class.
There user is domain object which cannot be injected, but UserViewer also needs UserConstants which is high-level object injected by IoC-container.
I want to inject UserConstants from the IoC-container, but I also need a transient runtime parameter User here.
What is wrong with the design?
Thanks in advance!
UPDATE
It seems I was not precise enough with my question. What I really need is an example how to do this:
create instance of class UserViewer(User user, UserService service), where user is passed as the parameter and service is injected from IoC.
If I inject UserViewer viewer then how do I pass user to it?
If I create UserViewer viewer manually then how do I pass service to it?
there's nothing wrong with this design. you use Factories for that, which have one leg in the domain, one leg in infrastructure.
You can either write them manually, or have the container do that for you, by things like TypedFactoryFacility in Windsor.
Also when your domain objects come from persistence layer you can plug your container there to inject the services they require (NHibernate can do that).
But what if my domain "fine-grained" object depends on some application-level object?
It is precisely this that is considered bad-practice. I would say the problems could be:
There are tons of these objects, so there can be performance and memory issues.
The POJO style is that they can be used in all environments (persisted in the database, processed in business algorithms and rules, read and set in view technologies, serialized and send over the network). Injecting application-level objects in them could cause the following problems:
In your architecture, you probably have the rule that some (most) application-level objects are usable in some layers, not in others. Because all layers have access to the pojos, the rule would be violated transitively.
When serialized and rebuild in another JVM, what would be the meaning of your application-level objects. They are useless, they must be changed for the local equivalents...
Typically, the pojos that constitute your domain are self-contained. They can have access to other pojos (and many enums), that's all.
In addition to the data, they have methods that implement the details of the business rules or algorithms (remember the OO idea of grouping data and code that work on it ;-) ):
This is especially good when they have inheritance, as this allow to customize a business rule for some pojo by providing a different implementation (differing case without if or switch: remember OO? ;-) ).
Any code that requires access to application-level objects (like accessing the database) is taken out, for example to a Service or Manager. But that code stays high level, thus readable and simple, because the pojos themselves take care of the low level details (and the special cases).
After the fact, you often find out that the pojo methods get reused a lot, and composed in different ways by the Services or Managers. That's a big win on reducing duplication, the methods names provide much needed "meaning", and provide an easier access to developers that are new to a module.
For your update:
create instance of class UserViewer(User user, UserService service), where user is passed as the parameter and service is injected from IoC.
If I inject UserViewer viewer then how do I pass user to it?
If I create UserViewer viewer manually then how do I pass service to it?
In that case, you need a factory method (possibly on a Factory or Locator of yours). It could look at follow, separating the two parts:
public UserViewer createUserViewer(User user) {
UserViewer viewer = instantiateBean(UserViewer.class);
viewer.setUser(user);
return viewer;
}
private <E> E instantiateBean(Class<E> clazz) {
// call the IoC container to create and inject a bean
}
Say I have a class called PermissionManager which should only exist once for my system and basically fulfills the function of managing various permissions for various actions in my application. Now I have some class in my application which needs to be able to check a certain permission in one of its methods. This class's constructor is currently public, i.e. used by API users.
Until a couple of weeks ago, I would have simply had my class call the following pseudo-code somewhere:
PermissionManager.getInstance().isReadPermissionEnabled(this)
But since I have noticed everyone here hating singletons + this kind of coupling, I was wondering what the better solution would be, since the arguments I have read against singletons seem to make sense (not testable, high coupling, etc.).
So should I actually require API users to pass in a PermissionManager instance in the constructor of the class? Even though I only want a single PermissionManager instance to exist for my application?
Or am I going about this all wrong and should have a non-public constructor and a factory somewhere which passes in the instance of PermissionManager for me?
Additional info Note that when I say "Dependency Injection", I'm talking about the DI Pattern...I am not using any DI framework like Guice or Spring. (...yet)
If you are using a dependency-injection framework, then the common way to handle this is to either pass in a PermissionsManager object in the constructor or to have a property of type PermissionsManager that the framework sets for you.
If this is not feasible, then having users get an instance of this class via factory is a good choice. In this case, the factory passes the PermissionManager in to the constructor when it creates the class. In your application start-up, you would create the single PermissionManager first, then create your factory, passing in the PermissionManager.
You are correct that it is normally unwieldy for the clients of a class to know where to find the correct PermissionManager instance and pass it in (or even to care about the fact that your class uses a PermissionManager).
One compromise solution I've seen is to give your class a property of type PermissionManager. If the property has been set (say, in a unit test), you use that instance, otherwise you use the singleton. Something like:
PermissionManager mManager = null;
public PermissionManager Permissions
{
if (mManager == null)
{
return mManager;
}
return PermissionManager.getInstance();
}
Of course, strictly speaking, your PermissionManager should implement some kind of IPermissionManager interface, and that's what your other class should reference so a dummy implementation can be substituted more easily during testing.
You can indeed start by injecting the PermissionManager. This will make your class more testable.
If this causes problems for the users of that class you can have them use a factory method or an abstract factory. Or you can add a parameterless constructor that for them to call that injects the PermissionManager while your tests use another constructor that you can use to mock the PermissionManager.
Decoupling your classes more makes your classes more flexible but it can also make them harder to use. It depends on the situation what you'll need. If you only have one PermissionManager and have no problem testing the classes that use it then there's no reason to use DI. If you want people to be able to add their own PermissionManager implementation then DI is the way to go.
If you are subscribing to the dependency injection way of doing things, whatever classes need your PermissionManager should have it injected as an object instance. The mechanism that controls its instantiation (to enforce the singleton nature) works at a higher level. If you use a dependency injection framework like Guice, it can do the enforcement work. If you are doing your object wiring by hand, dependency injection favors grouping code that does instantiation (new operator work) away from your business logic.
Either way, though, the classic "capital-S" Singleton is generally seen as an anti-pattern in the context of dependency injection.
These posts have been insightful for me in the past:
Using Dependency Injection to Avoid Singletons
How to Think About the "new" Operator with Respect to Unit Testing
So should I actually require API users to pass in a PermissionManager instance in the constructor of the class? Even though I only want a single PermissionManager instance to exist for my application?
Yes, this is all you need to do. Whether a dependency is a singleton / per request / per thread or a factory method is the responsibility of your container and configuration. In the .net world we would ideally have the dependency on an IPermissionsManager interface to further reduce coupling, I assume this is best practice in Java too.
The singleton pattern is not bad by itself, what makes it ugly is the way it's commonly used, as being the requirement of only wanting a single instance of a certain class, which I think it's a big mistake.
In this case I'd make PermissionManager a static class unless for any reason you need it to be an instanciable type.