Autofac finds "best" ctor when registering a component like this
builder.RegisterType<MyType>()...;
or
builder.RegisterType(typeof(MyType))...;
I'm looking for a way to reuse Autofac's reflection code to find ctor parameters. Is this logic available in some public API?
The reason I want above...
I have some components with unknown types at compile time (=dynamic proxies). Currently I register them like this
builder.Register(c =>
{
var dep1 = c.Resolve<IFoo>();
var dep2 = c.Resolve<IBar>();
var dep3 = c.Resolve<IFooBar>();
return someProxyFactory.CreateProxyFrom<MyType>(dep1, dep2, dep3);
}...;
--- Edit ---
The same question is valid if no proxy is involved but when RegisterType cannot be used. Eg...
builder.Register(c =>
{
[...]
if(something)
return new SomeType(dep1, dep2, dep3);
else
return new SomeOtherType(dep1, dep4, dep2, dep5);
}
Here I also would like to reuse Autofac's "find ctor logic" if possible.
--- End edit ---
It works fine but, if possible, I would like to use autofac's logic to find the ctor dependencies for me. I want to write something like
builder.Register(c =>
{
object[] ctorDependencies = letAutofacDoTheSimilarWorkAsInRegisterType(typeof(MyType));
return someProxyFactory.Create<MyType>(ctorDependencies);
}
Is this possible or do I have to write my own logic for this? Or is some completely different approach available for this scenario?
Not quite the answer you're after I guess, but could you use the existing DynamicProxy2? This enables you to attach interceptors to registered interfaces, much like what you are doing with your proxy factory.
Update: you could possibly use the IConstructorFinder and IConstructorSelector interfaces in the Autofac.Core.Activators.Reflection namespace, implemented by the PublicConstructorFinder and MostParametersConstructorSelector respectively.
Related
I am still fairly new to Autofac and Nlog and I need some help in understanding what is taking place in my Autofac LoggingModule for Nlog. It works as expected thanks to following the injecting-nlog-with-autofacs-registergeneric. But rather than just copy paste, I would like to make sure I understand what is occurring in each method (Load & AttachToComponentRegistration). If you could review my thoughts and further clarify anything I have incorrect (quite a bit I am sure), I would greatly appreciate it. Thank you in advance!
Database Target using Nlog
Dependency Injection using Autofac
ASP.NET MVC web app for learning
Dvd Libary app (DvdAdd, DvdEdit, DvdDelete, DvdList)
LoggingModule
public class LoggingModule : Module
{
protected override void Load(ContainerBuilder builder)
{
builder
.Register((c, p) => new LogService(p.TypedAs<Type>()))
.AsImplementedInterfaces();
}
protected override void AttachToComponentRegistration(IComponentRegistry componentRegistry, IComponentRegistration registration)
{
registration.Preparing +=
(sender, args) =>
{
var forType = args.Component.Activator.LimitType;
var logParameter = new ResolvedParameter(
(p, c) => p.ParameterType == typeof(ILog),
(p, c) => c.Resolve<ILog>(TypedParameter.From(forType)));
args.Parameters = args.Parameters.Union(new[] { logParameter });
};
}
}
My understanding of the code within Load()
c - The parameter c, provided to the expression, is the component context(an IComponentContext object) in which the component is being created. The context in which a service can be accessed or a component's dependencies resolved.
p - An IEnumerable with the incoming parameter set
AsImplementedInterfaces - Autofac allows its users to register the types explicitly or implicitly. While "As" is used for explicit registrations, "AsImplementedInterfaces" and "AsSelf" are used for implicit ones. In other words, the container automatically registers the implementation against all the interfaces it implements.
Thoughts: The Load method code registers a new LogService class (which represents "c") with the type of logger (which represents "p") as the constructor parameter for the LogService class
Questions:
Are my thoughts above correct?
Should it be SingleInstance or should it / will it only live as long as the calling classes scope? (I am thinking about my Unit Of Work)
My understanding of the code within AttachToComponentRegistration()
AttachToComponentRegistration method - Override to attach module-specific functionality to a component registration.
AttachToComponentRegistration Parameters:
IComponentRegistry componentRegistry - Provides component registrations according to the services they provide.
IComponentRegistration registration - Describes a logical component within the container.
registration.Preparing - Fired when a new instance is required. The instance can be provided in order to skip the regular activator, by setting the Instance property in the provided event arguments.
var forType = args.Component.Activator.LimitType;
args = Autofac.Core.PreparingEventArgs - Fired before the activation process to allow parameters to be changed or an alternative instance to be provided.
Component = PreparingEventArgs.Component Property - Gets the component providing the instance being activated
Activator = IComponentRegistration.Activator Property - Gets the activator used to create instances.
LimitType = IInstanceActivator.LimitType Property - Gets the most specific type that the component instances are known to be castable to.
Thoughts on forType - As I understand it, this variable holds the Name and FullName of the calling class from where the logging service is being called?
forType Debugger Image
Questions:
Are my thoughts forType correct?
var logParameter = new ResolvedParameter(
(p, c) => p.ParameterType == typeof(ILog),
(p, c) => c.Resolve<ILog>(TypedParameter.From(forType)));
ResolvedParameter - can be used as a way to supply values dynamically retrieved from the container,
e.g. by resolving a service by name.
Thoughts on logParameter - This is where I start to get lost. So does, it check that the Parameter is of Type ILog and if so it will then resolve it with the constructor parameter and pass in forType variable?
Questions:
Are my thoughts on logParameter above correct?
args.Parameters = args.Parameters.Union(new[] { logParameter });
args.Parameters = PreparingEventArgs.Parameters Property - Gets or sets the parameters supplied to the activator.
args.Parameters.Union = Produces the set union of two sequences by using the default equality comparer. Returns an System.Collections.Generic.IEnumerable`1 that contains the elements from both input sequences, excluding duplicates.
Thoughts on args.Parameters - I really do not know at this point other than to guess that it returns a collection of Parameters and removes duplicates?
Questions:
Could you help talk me through what is going on in args.Parameters?
logParameter Debugger Image
Nlog Database Table Image
LogService class
public class LogService : ILog
{
private readonly ILogger _log;
public LogService(Type type)
{
_log = LogManager.GetLogger(type.FullName);
}
public void Debug(string message, params object[] args)
{
Log(LogLevel.Debug, message, args);
}
public void Info(string message, params object[] args)
{
Log(LogLevel.Info, message, args);
}
public void Warn(string message, params object[] args)
{
Log(LogLevel.Warn, message, args);
}
public void Error(string message, params object[] args)
{
Log(LogLevel.Error, message, args);
}
public void Error(Exception ex)
{
Log(LogLevel.Error, null, null, ex);
}
public void Error(Exception ex, string message, params object[] args)
{
Log(LogLevel.Error, message, args, ex);
}
public void Fatal(Exception ex, string message, params object[] args)
{
Log(LogLevel.Fatal, message, args, ex);
}
private void Log(LogLevel level, string message, object[] args)
{
_log.Log(typeof(LogService), new LogEventInfo(level, _log.Name, null, message, args));
}
private void Log(LogLevel level, string message, object[] args, Exception ex)
{
_log.Log(typeof(LogService), new LogEventInfo(level, _log.Name, null, message, args, ex));
}
}
ILog interface
public interface ILog
{
void Debug(string message, params object[] args);
void Info(string message, params object[] args);
void Warn(string message, params object[] args);
void Error(string message, params object[] args);
void Error(Exception ex);
void Error(Exception ex, string message, params object[] args);
void Fatal(Exception ex, string message, params object[] args);
}
There's a lot to unpack here. You're not really asking for an answer to a specific question so much as a code walkthrough and explanation of an existing solution that works, so I might suggest posting to StackExchange Code Review if you need much more than what I'm going to give you here. Not trying to be unhelpful, but, like, if your question is, "Is my thinking right?" and the answer is "sort of," there's a lot of discussion on each individual point to explain why "sort of" is the answer (or "no," or "yes," as the case may be). It can turn into a lengthy answer, followed up by additional questions for clarification, which require yet additional answers... and StackOverflow isn't really a discussion forum capable of those sorts of things.
[i.e., I'll take probably an hour and write up an answer here... but I can't promise I'll actually be back to follow up on anything because there are other questions to answer and other things I need to allocate time to. StackOverflow is really more about "How do I...?" or other things that have a single, reasonably concrete answer.]
First, I recommend diving in yourself with a debugger on some breakpoints to actually see what's going on. For example, you asked what's in LimitType in one area - you could pretty easily answer that one by just sticking a breakpoint on that line and looking at the value. This will be a good way to follow up for additional clarification yourself - breakpoints for the win.
Second, I recommend spending some time with the Autofac docs. There's a lot of documentation out there that can answer questions.
The NLog module here appears to be based on the log4net module in the documentation which has a bit more explanation of what's going on.
There's an explanation of parameters (like TypedParameter) and how they're used.
Given the docs can round out some things that may not be clear, rather than try to address each "are my thoughts correct" item, let me just heavily annotate the module and hope that clarifies things.
// General module documentation is here:
// https://autofac.readthedocs.io/en/latest/configuration/modules.html
public class LoggingModule : Module
{
// Load basically registers types with the container just like
// if you were doing it yourself on the ContainerBuilder. It's
// just a nice way of packaging up a set of registrations so
// they're not all in your program's "Main" method or whatever.
protected override void Load(ContainerBuilder builder)
{
// This is a lambda registration. Docs here:
// https://autofac.readthedocs.io/en/latest/register/registration.html#lambda-expression-components
// This one uses both the component context (c) and the incoming
// set of parameters (p). In this lambda, the parameters are NOT the set of constructor
// parameters that Autofac has resolved - they're ONLY things that
// were MANUALLY specified. In this case, it's assuming a TypedParameter
// with a System.Type value is being provided manually. It's not going
// to try resolving that value from the container. This is going hand-in-hand
// with the logParameter you see in AttachToComponentRegistration.
// Parameter docs are here:
// https://autofac.readthedocs.io/en/latest/resolve/parameters.html
// In general if you resolve something that has both manually specified parameters
// and things that can be resolved by Autofac, the manually specified parameters
// will take precedence. However, in this lambda it's very specifically looking
// for a manually specified parameter.
// You'll want to keep this as a default InstancePerDependency because you probably
// want this to live as long as the thing using it and no longer. Likely
// NLog already has object pooling and caching built in so this isn't as
// expensive as you think, but I'm no NLog expert. log4net does handle
// that for you.
builder
.Register((c, p) => new LogService(p.TypedAs<Type>()))
.AsImplementedInterfaces();
}
// This method attaches a behavior (in this case, an event handler) to every
// component registered in the container. Think of it as a way to run a sort
// of "global foreach" over everything registered.
protected override void AttachToComponentRegistration(
IComponentRegistry componentRegistry,
IComponentRegistration registration)
{
// The Preparing event is called any time a new instance is needed. There
// are docs for the lifetime events but Preparing isn't on there. Here are the
// docs and the issue I filed on your behalf to get Preparing documented.
// https://autofac.readthedocs.io/en/latest/lifetime/events.html
// https://github.com/autofac/Documentation/issues/69
// You can see the Preparing event here:
// https://github.com/autofac/Autofac/blob/6dde84e5b0a3f82136a0567a84da498b04e1fa2d/src/Autofac/Core/IComponentRegistration.cs#L83
// and the event args here:
// https://github.com/autofac/Autofac/blob/6dde84e5b0/src/Autofac/Core/PreparingEventArgs.cs
registration.Preparing +=
(sender, args) =>
{
// The Component is the thing being resolved - the thing that
// needs a LogService injected. The Component.Activator is the
// thing that is actually going to execute to "new up" an instance
// of the Component. The Component.Activator.LimitType is the actual
// System.Type of the thing being resolved.
var forType = args.Component.Activator.LimitType;
// The docs above explain ResolvedParameter - basically a manually
// passed in parameter that can execute some logic to determine if
// it satisfies a constructor or property dependency. The point of
// this particular parameter is to provide an ILog to anything being
// resolved that happens to have an ILog constructor parameter.
var logParameter = new ResolvedParameter(
// p is the System.Reflection.ParameterInfo that describes the
// constructor parameter that needs injecting. c is the IComponentContext
// in which the resolution is being done (not used here). If this
// method evaluates to true then this parameter will be used; if not,
// it will refuse to provide a value. In this case, if the parameter
// being injected is an ILog, this ResolvedParameter will tell Autofac
// it can provide a value.
(p, c) => p.ParameterType == typeof(ILog),
// p and c are the same here, but this time they're used to actually
// generate the value of the parameter - the ILog instance that should
// be injected. Again, this will only run if the above predicate evaluates
// to true. This creates an ILog by manually resolving from the same
// component context (the same lifetime scope) as the thing that
// needs the ILog. Remember earlier that call to p.AsTyped<Type>()
// to get a parameter? The TypedParameter thing here is how that
// value gets poked in up there. This Resolve call will effectively
// end up calling the lambda registration.
(p, c) => c.Resolve<ILog>(TypedParameter.From(forType)));
// The thing being resolved (the component that consumes ILog) now
// needs to be told to make use of the log parameter, so add it into
// the list of parameters that can be used when resolving that thing.
// If there's an ILog, Autofac will use this specified parameter to
// fulfill the requirement.
args.Parameters = args.Parameters.Union(new[] { logParameter });
};
}
}
Something missing from this that's present in the log4net module example is the ability to do property injection for the logger. However, I'm not going to solve that here; you can look at the example right in the documentation and take that as an exercise to work on if you need that functionality.
I hope that helps. I'll probably not be coming back to follow up on additional questions, so if this isn't enough, I very, very much recommend setting some breakpoints, maybe setting up some tiny minimal-repro unit tests, that sort of thing, and do some deeper exploration to get clarity. Honestly, it's one thing to have someone else explain it, but it's another to actually see it in action and dive into the source of various projects. You'll come out with a fuller understanding with the latter approach, even if it's potentially not as fast.
If you want to have another middleware/object in a middleware
you have to use a factory like
namespace App\Somnething;
use Interop\Container\ContainerInterface;
class MyMiddlewareFactory
{
public function __invoke(ContainerInterface $container, $requestedName)
{
return new $requestedName(
$container->get(\App\Path\To\My\Middleware::class)
);
}
}
So MyMiddleware would have been injected with \App\Path\To\My\Middleware and we would be able to access it.
Question:
would it be wrong to inject the middleware with the app itself or the container? Like:
namespace App\Somnething;
use Interop\Container\ContainerInterface;
use Zend\Expressive\Application;
class MyMiddlewareFactory
{
public function __invoke(ContainerInterface $container, $requestedName)
{
return new $requestedName(
$container->get(Application::class)
);
}
}
This way it would be possible to get anything ~on the fly.
Like
namespace App\Somnething;
use Zend\Expressive\Application;
class MyMiddleware
{
/** #var Application $app */
protected $app;
public function __construct(Application $app)
{
$this->app = $app;
}
public function __invoke($some, $thing)
{
if ($some and $thing) {
$ever = $this->app
->getContainer()
->get(\Path\To\What\Ever::class);
$ever->doSome();
}
}
}
You don't inject middleware into other middleware. You inject dependencies like services or repositories. Each middleware takes care of a specifc task like authentication, authorization, localization negotiation, etc. They are executed one after the other. They mess around with the request and pass the request to the next middleware. Once the middleware stack has been exhausted, the response is returned all the way back through all the middleware in reverse order until it finally reaches the outer layer which displays the output. You can find a flow overview in the expressive docs.
I wouldn't advice to inject the container and certainly not the app itself. Although it might be easy during development, your application becomes untestable. If you inject only the services that are needed into a middleware, action or service you can easily mock those during tests. After a while you get used to writing factories where needed and it goes pretty fast.
The same goes for injecting the entity manager (if you use doctrine). It's easier to test an application if you only inject the needed repositories, which you can easily mock.
Having said this, if you are looking for an easy way to inject dependencies, zend-servicemanager can do that. Take a look at abstract factories. With an abstract factory you can create one factory for all your action classes:
<?php
namespace App\Action;
use Interop\Container\ContainerInterface;
use ReflectionClass;
use Zend\ServiceManager\Factory\AbstractFactoryInterface;
class AbstractActionFactory implements AbstractFactoryInterface
{
public function __invoke(ContainerInterface $container, $requestedName, array $options = null)
{
// Construct a new ReflectionClass object for the requested action
$reflection = new ReflectionClass($requestedName);
// Get the constructor
$constructor = $reflection->getConstructor();
if (is_null($constructor)) {
// There is no constructor, just return a new class
return new $requestedName;
}
// Get the parameters
$parameters = $constructor->getParameters();
$dependencies = [];
foreach ($parameters as $parameter) {
// Get the parameter class
$class = $parameter->getClass();
// Get the class from the container
$dependencies[] = $container->get($class->getName());
}
// Return the requested class and inject its dependencies
return $reflection->newInstanceArgs($dependencies);
}
public function canCreate(ContainerInterface $container, $requestedName)
{
// Only accept Action classes
if (substr($requestedName, -6) == 'Action') {
return true;
}
return false;
}
}
I wrote a blog post about that.
At the end of the day it's your own decision, but best practice is not injecting the app, container or the entity manager. It will make your life easier if you need to debug your middleware and / or write tests for it.
Injecting the application or container in your middleware is possible but it is not good idea at all:
1) Inversion Of Control (IoC)
It violated the inversion of control principle, your class must not have any knowledge about the IoC container.
2) Dependency Inversion Principle (DIP)
Dependency inversion principle states that "high-level modules should not depend on low-level modules", so your higher level middleware class depends on the infrastructure/framework.
3) Law of Demeter (LoD)
According to the law of demeter, the unit should have limited knowledge about other units, it should know only about its closely related units.
The MyMiddleware::class has too much knowledge other units, first of all, it knows about the Application::class, then it knows that the Application knows about the Container, then it knows that the Container knows about the What\Ever::class and so on.
This kind of code violates some of the most important OOP principles, leads to horrible coupling with the framework, it has implicit dependencies and least but not last, it is hard to be read and understood.
I have this code in my controller, it takes 'procedure_type' from the request and checks to see if a ProcedureType with that name exists. If it does it uses the object, if not it creates a new ProcedureType, then return the new object to use.
// Check the typed in ProcedureType against existing types.
$procedureTypes = $entityManager->getRepository('IncompassSurgeryBundle:ProcedureType')->findBy(array('name' => $request->request->get('procedure_type'), 'vendor' => $vendorId));
if (empty($procedureTypes)) {
// Create Procedure Type
$procedureType = new ProcedureType();
$procedureType->setVendor($vendor)
->setName($request->request->get('procedure_type'))
->setCreated(new \DateTime())
->setUpdated($procedureType->getCreated());
$entityManager->persist($procedureType);
} else {
$procedureType = $procedureTypes[0];
}
I don't think this is the best way to do this, I'd like to move the code into a function, say checkProcedureType(), but I don't know where the best place is to put that. I don't think it could go in the Entity or Repository classes, and moving it to a private function in the controller doesn't feel right.
I'm sure there is a class type that I'm not aware of, that extends the Entity. Or maybe I should just put these functions in my entity classes.
Service are the answer to almost everything in Symfony 2. Create a service like this :
namespace Your\Bundle\Service;
class ProcedureService // Call this the way you want
{
protected $entityManager;
public function __construct($entityManager)
{
$this->entityManager = $entityManager;
}
public function callMeTheWayYouWant($vendorId, $vendor)
{
// Check the typed in ProcedureType against existing types.
$procedureTypes = $this->entityManager->getRepository('IncompassSurgeryBundle:ProcedureType')->findBy(array('name' => $request->request->get('procedure_type'), 'vendor' => $vendorId));
if (empty($procedureTypes)) {
// Create Procedure Type
$procedureType = new ProcedureType();
$procedureType->setVendor($vendor)
->setName($request->request->get('procedure_type'))
->setCreated(new \DateTime())
->setUpdated($procedureType->getCreated());
$this->entityManager->persist($procedureType);
} else {
$procedureType = $procedureTypes[0];
}
// The rest of your code
}
}
In your services.yml file :
your_service:
class: Your\Bundle\Service\ProcedureService
arguments: [#doctrine.orm.entity_manager]
Then use it in your controller :
$this->get('your_service')->callMeTheWayYouWant($vendorId, $vendor);
If logic is somehow related to acessing database I always go for repository. However, if cases like yours, I tend to analyze it's dependency map.
Does your code repeats in some other method within same class, only?
If so, go for private method.
Is this part of code reused somewhere else but does not rely on some services?
You could externalize logic by creating separate class and static method which executes the code. Beware: Tends to get messy really quick
Finally, does your code rely on services/configuration?
Create a separate service, inject the services/configuration and invoke it's method. Adds a bit of overhead, if your abuse it, but you should be fine
Personally, in your example, I would go for private method, but that's just my opinion.
I have the following registration
builder.Register<Func<Type, IRequestHandler>>(
c => request => (IRequestHandler)c.Resolve(request));
Basically I am trying to register a factory method that resolves an instance of IRequestHandler from a given type.
This works fine until the version 2.4.3.700. But now I am getting a the following error..
Cannot access a disposed object.
Object name: 'This resolve operation has already ended. When
registering components using lambdas,
the IComponentContext 'c' parameter to
the lambda cannot be stored. Instead,
either resolve IComponentContext again
from 'c', or resolve a Func<> based
factory to create subsequent
components from.'.
UPDATE
I was trying to limit autofac's exposure to the rest of the projects in the solution. Nick, thanks for the hint, now my registration looks like this...
builder.Register<Func<Type,IRequestHandler>>(c =>
{
var handlers = c.Resolve<IIndex<Type,RequestHandler>>();
return request => handlers[request];
});
The c in this expression is a temporary, so this code while previously functional, is broken. Autofac 2.4.5 detects this problem while earlier versions silently ignored it.
To fix the issue, explicitly resolve IComponentContext:
builder.Register<Func<Type, IRequestHandler>>(c => {
var ctx = c.Resolve<IComponentContext>();
return request => (IRequestHandler)ctx.Resolve(request));
});
The functionality you're emulating here might be better represented using keys and indexes, e.g. see Interrupted chain of IoC or http://code.google.com/p/autofac/wiki/TypedNamedAndKeyedServices.
I had a similar problem as the user6130. I wanted to avoid using IIndex in my class implementation and pass in a service resolver into my constructor instead.
So now I have my service implementation with the following constructor:
public MvcMailer(Converter<string, MailerBase> mailerResolver)
{
_resolver = mailerResolver;
}
I wanted to used keyed services without directly relying on the Autofac namespace. I was getting the same error until I restructured the configuration as such.
1) Scan for all my mailer implementations and index via class name (could be improved)
builder.RegisterAssemblyTypes(System.Reflection.Assembly.GetExecutingAssembly())
.Where(t => t.Name.EndsWith("Mailer")).Keyed<Mvc.Mailer.MailerBase>(t => t.Name.Replace("Mailer", "").ToLower());
2) Register the converter in Autofac config
builder.Register<System.Converter<string,Mvc.Mailer.MailerBase>>(c => {
var all = c.Resolve<Autofac.Features.Indexed.IIndex<string,Mvc.Mailer.MailerBase>>();
return delegate(string key)
{
return all[key];
};
});
3) Register like other types of components and let Autofac handle the Converter injection
builder.RegisterType<Mailers.MvcMailer>().As<Mailers.IMailer>();
I have an application which does data processing. There is
class Pipeline {
IEnumerable<IFilter> Filters {get; set;}
I register filters implementations as
builder.RegisterType<DiversityFilter>().As<IFilter>();
builder.RegisterType<OverflowFilter>().As<IFilter>();
...
So far so good. Now, for experimentation and fine-tuning I want to be able to override any filter implementation in config file with a program(script) which would read data from stdin, process it and send data to stdout. I've implemented a module with "fileName", "args" and "insteadOf" custom properties, described module in xml and got it called.
In the module I register my "ExecutableFilter" but how do I make it run "instead of" desired service? If I try do it like this:
builder.RegisterType<ExecutableFilter>().As<DiversityFilter>()
then I get an exception " The type 'ExecutableFilter' is not assignable to service 'DiversityFilter'.". Ok, this is logical. But what are my options then?
Once you've overridden the registration for IFilter "After" with your wire-tap, you won't be able to resolve it from the container, as the new registration will be activated instead, hence the circular lookup.
Instead, create and register a module that hooks into the filter's creation, and replaces the instance with the 'wire tapped' one:
class WiretapModule : Module
{
override void AttachToComponentRegistration(
IComponentRegistration registration,
IComponentRegistry registry)
{
if (registration.Services.OfType<KeyedService>().Any(
s => s.ServiceKey == After && s.ServiceType == typeof(IFilter)))
{
registration.Activating += (s, e) => {
e.Instance = new WireTap((IFilter)e.Instance, new ExecuteProvider(fileName, args))
};
}
}
}
(Cross-posted to the Autofac group: https://groups.google.com/forum/#!topic/autofac/yLbTeuCObrU)
What you describe is part container work, part business logic. The challenge is to keep separation of concerns here. IMO, the container should do what it is supposed to do, that is building and serving up instances or collections thereof. It should not do the "instead of" in this case. I would rather "enrich" the services with enough information so that the pipeline make the decision.
The "enrichment" can be accomplished by making the ExecutableFilter implement a more distinct interface.
interface IInsteadOfFilter : IFilter { }
...
builder.RegisterType<ExecutableFilter>().As<IFilter>();
...
class Pipeline
{
IEnumerable<IFilter> Filters {get;set;}
public void DoTheFiltering()
{
var filters = Filters.OfType<IInsteadOfFilter>();
if (!insteadof.Any())
filters = Filters;
foreach(var filter in filters)
{
...
}
}
You could also solve this using the metadata infrastructure, which gives us an even more expressive way of differentiating services.