I'm currently building a CQS-style DDD-application. I'm having some doubts on how all 'components' work with each other.
But first I'll give a brief overview about the application's structure:
ApplicationService
-> Receives command objects
-> doesn't return any results
-> Acts on Domain model
-> Speaks with Aggregate repository for domain modifications
QueryService
-> Bypasses domain model; doesn't speak with Aggregate Repositories
-> Executes queries against database to populate view
-> Returns 'Representation' objects
REST Controller
-> Receives HTTP requests and binds 'body content' & request params to Command objects
-> delegates to ApplicationService for POST, PUT & DELETE requests
-> Always returns at least some HTTP code
-> delegates to QueryService for GET requests
Infrastructure
-> Handles persistence to DB
-> Contains some scheduling operations
-> Handles foreign domain events our domain model is 'interested' in
'Open Host'
-> This is mainly a Facade to be used by other domains
-> Facade delegates methods to ApplicationService for domain modifications and to QueryService for data retrieval (bypassing Repositories)
My Questions:
Is it OK that a DomainEventHandler corresponds with a Repository and invokes some methods on a Aggregate? Or should it always correspond with an ApplicationService?
QueryService returns 'Representation' objects. These are used by UI AND by 'Open Host' Facade as return value. Is it OK these objects are reused as return value by Facade? Or should Facade create their own Objects, even the results are basically the same?
ApplicationService takes 'Commands' as input parameters. Is it OK these Commands are also used by the Open Host Facade? Or should the Facade only accept primitive values and convert them to Commands when delegating to ApplicationService?
DomainEventHandlers seem to reside on 'Infrastructure' layer. Is it possible that an ApplicationService or Domain Service also subscribes to an Domain Event? Or is this always an Infrastructure responsibility?
All advice is very welcome!
Is it OK that a DomainEventHandler corresponds with a Repository and invokes some methods on a Aggregate? Or should it always correspond with an ApplicationService?
In my experience, any handlers are application services.
QueryService returns 'Representation' objects. These are used by UI AND by 'Open Host' Facade as return value. Is it OK these objects are reused as return value by Facade? Or should Facade create their own Objects, even the results are basically the same?
There is a lot of discussion here about the differences between Open Host service and Application Service. It is not clear to me who would be using Open Host service, or why it exists.
ApplicationService takes 'Commands' as input parameters. Is it OK these Commands are also used by the Open Host Facade? Or should the Facade only accept primitive values and convert them to Commands when delegating to ApplicationService?
I would pass in primitives on the edges of the application and convert them into commands which are then handled in the Application Services
DomainEventHandlers seem to reside on 'Infrastructure' layer. Is it possible that an ApplicationService or Domain Service also subscribes to an Domain Event? Or is this always an Infrastructure responsibility?
I've always considered my handlers to be Application Services - things that are responsible for orchestrating a user case. So the use case might be "when EventX is received, send an email and update the database". In this example, you would probably consider "the code that sends the email" and "the code that saves to the database" to be infrastructure concerns, but the handler itself would not be.
public class ExampleHandler : IHandle<ExampleEvent>
{
private IRepository _repo;
private ISendEmails _emailer;
public ExampleHandler(Repository repo, ISendEmails emailer)
{
.... set the private fields..
}
public void When(ExampleEvent event)
{
_emailer.Send(event.whatever);
_repo.Save(something);
}
}
To be honest, I don't really think in terms of layers - i prefer a hexagonal architecture style of thinking. In the above example, the event handlers would just have dependencies injected into them and then go about their business.
Related
I am thinking about scenario in a way of Domain Driven design, where I have entity, lets say Cv (Curriculum vitae), which state is saved in database via repository.
Now I need to store part of the Cv in another system (ElasticSearch), which is crucial for whole app functionality like searching.
How to handle it? I am thinking about these 2 options:
1. Use domain service IndexCvGatewayServiceInterface (as interfaces implemented in infrastructure)
class CvEntity
{
public function approve(CvRepositoryInterface $cvRepository, IndexCvGatewayServiceInterface $indexService)
{
$cvRepository->update($this);
$indexService->update($this);
}
}
2. Listen to domain event (create infrastructure listener)
class CvEntity
{
public function approve(CvRepositoryInterface $cvRepository, EventDispatcheInterface $dispatcher)
{
$cvRepository->update($this);
$dispatcher->dispatch(new CvApprovedEvent($this));
}
}
I like option 2. because it separates logic for non state change purposes into infrastructure, but there is also concern, that we should know about searching as important part of our app.
You're facing here Write and Read model. Ideally after persist your entity/aggregate in the write model you should dispatch the uncommitted events of this entity and listing/subscribe to them to generate the projections (partials in elastic in your use case). For reference: https://github.com/jorge07/symfony-5-es-cqrs-boilerplate/blob/symfony-5/src/Infrastructure/User/ReadModel/Projections/UserProjectionFactory.php#L17
IMO, Entity should not contain the repository.
I've read about the Double Dispatch pattern, which enables to pass service interfaces into aggregate methods: https://lostechies.com/jimmybogard/2010/03/30/strengthening-your-domain-the-double-dispatch-pattern/, http://blog.jonathanoliver.com/dddd-double-dispatch/.
In my domain I have a BitbucketIntegration aggregate, which is local copy of a remote bitbucket account with some additional domain specific data. Now, I have to synchronize repositories and teams, etc.. from the cloud to be able to do business operations on them. In my first implementation I was using a service to access the Bitbucket Cloud, then set the aggregate's repositories, teams, account. This way I had a DDD mixed with Anemic Domain Model, since half of the aggregates state was set using setter-like methods from the service. With Double Dispatch I can pass e.g. a BitbucketService interface into method arguments. This way, the aggregate can protect it's invariants more, since some of the data can only be verified by connecting to the rest service (e.g. if the aggregate's accessToken, bitbucketAccount and repositories are in sync), which was the service's responsibility. One more thing that smells is that I have an accessToken field in my aggregate, which is only a technical concern.
Are there any recommended patterns for keeping a copy of a remote resource in a ddd aggregate? Also, how to keep the technical side out of it? Or was the first method with a domain service good enough?
Now the code looks something like:
class BitbucketIntegration extends Aggregate<UUID> {
accountId: BitbucketId
repos: List<Repository>
localData: ...
// ... and more
Single integrateWith(accessToken, queryService) {
var id = queryService.getAccountAsync(accessToken);
var repos = queryService.getReposAsync(accessToken);
return Single.zip(id, repos,
(i, r) -> new BitbucketIntegratedEvent(accessToken, i, r))
.onSubscribe(event -> apply(event))
}
Observable doSomeBusinessLocally(data) { ... return events; }
// this is triggered by a saga
Single pollForChanges(queryService) {
var dataFromRemote = queryService.synchronizeAsync(this.accessToken);
....
return event;
}
}
class CommandHandler {
queryService: BitbucketService
Completable handle(integrateCmd) {
aggregate = repo.get(integrateCmd.id);
return aggregate.integrateWith(integrateCmd.accessToken, queryService)
.flatMap(event -> repo.store(event));
}
}
As a side note, I only query Bitbucket.
EDIT:
Martin Fowler writes about accessing an external system, including the definition of an Anti-Corruption Layer, which translates the remote resource representation to domain types.
If you inject infrastructure services into your Aggregate (by constructor or by method invocation) then you won't have a pure domain model anymore. This includes even services that have interfaces defined in the domain layer. It affects testability and introduces a dependency on the infrastructure. It also breaks the Single responsibility principle and it forces the Aggregate to know things it does not really need to.
The solution to this is to call the service before and pass the result to the Aggregate's method (i.e. in the Application layer).
I'm curious why the decision was made to couple the Service base class in ServiceStack to data access (via the Db property)? With web services it is very popular to use a Data Repository pattern to fetch the raw data from the database. These data repositories can be used by many services without having to call a service class.
For example, let's say I am supporting a large retail chain that operates across the nation. There are a number of settings that will differ across all stores like tax rates. Each call to one of the web services will need these settings for domain logic. In a repository pattern I would simply create a data access class whose sole responsibility is to return these settings. However in ServiceStack I am exposing these settings as a Service (which it needs to be as well). In my service call the first thing I end up doing is newing up the Setting service and using it inside my other service. Is this the intention? Since the services return an object I have to cast the result to the typed service result.
ServiceStack convenience ADO.NET IDbConnection Db property allows you to quickly create Database driven services (i.e. the most popular kind) without the overhead and boilerplate of creating a repository if preferred. As ServiceStack Services are already testable and the DTO pattern provides a clean endpoint agnostic Web Service interface, there's often not a lot of value in wrapping and proxying "one-off" data-access into a separate repository.
But at the same time there's nothing forcing you to use the base.Db property, (which has no effect if unused). The Unit Testing Example on the wiki shows an example of using either base.Db or Repository pattern:
public class SimpleService : Service
{
public IRockstarRepository RockstarRepository { get; set; }
public List<Rockstar> Get(FindRockstars request)
{
return request.Aged.HasValue
? Db.Select<Rockstar>(q => q.Age == request.Aged.Value)
: Db.Select<Rockstar>();
}
public RockstarStatus Get(GetStatus request)
{
var rockstar = RockstarRepository.GetByLastName(request.LastName);
if (rockstar == null)
throw HttpError.NotFound("'{0}' is no Rockstar".Fmt(request.LastName));
var status = new RockstarStatus
{
Alive = RockstarRepository.IsAlive(request.LastName)
}.PopulateWith(rockstar); //Populates with matching fields
return status;
}
}
Note: Returning an object or a strong-typed DTO response like RockstarStatus have the same effect in ServiceStack, so if preferred you can return a strong typed response and avoid any casting.
1) In most cases each Aggregate Root should define its own transactional boundary, in which case we don't need to expose IUnitOfWork interface in Domain Layer.
a) I assume in this situation a good option would be for a repository ( used by aggregate to enforce invariants applied within it ) to contain its very own instance of UoW ( if using EF, then this UoW instance could simply be of type DbContext )?
2)
a) But if for whatever reason transaction spans several aggregates ( thus more than one aggregate needs to be changed at one time ), then won't Domain Layer also need to contain IUnitOfWork interface?
b) Won't exposing IUnitOfWork interface in Domain Layer violate persistence ignorance rule?
c) If yes to b), doesn't then exposing IUnitOfWork defeat the purpose of having repositories?
Replying to Alexey Raga:
1)
I would advice against exposing repositories to aggregates. Repositories are there to give you aggregates, that's it.
a) Though I assume that majority of ddd architects don't have a problem with exposing repos to aggregates ( I'm only asking because I read several articles on repos and DDD and the impression I got is that authors ain't against exposing repos to aggregates - but now I'm not so sure anymore )?
b) So you're also against exposing repositories to domain services?
c) Judging by your answer I'm guessing that you consider exposing IUnitOfWork as a violation of PI?
2)Note that although my command handler (app service in a way)...
Do you normally implement command handlers as app services?
3)
public void Handle(ApproveOrderCommand command)
{
var order = Repository.Get(command.OrderId);
property.Approve(command.Comment, ServiceRequiredForOrderApproval);
Repository.Save(order);
}
Is property.Approve(...) a typo and you actually meant order.Approve(...)?
Thanx in advance
I would advice against exposing repositories to aggregates. Repositories are there to give you aggregates, that's it.
Look at it at that way: your domain is a "bubble" which only understands its own stuff. Meaning, it only understand its own value objects, domain services interfaces it declares, etc. I wouldn't include repositories in this set.
When your domain (an aggregate) needs something it should explicitly expose the dependency of what it needs, not just ask for some repository.
Services is what brings things together.
For example, my command handler could look like:
public class ApproveOrderCommandHandler : IHandle<ApproveOrderCommand>
{
//this might be set by a DI container, or passed to a constructor
public IOrderRepository Repository { get; set; }
public ISomeFancyDomainService ServiceRequiredForOrderApproval { get; set; }
public void Handle(ApproveOrderCommand command)
{
var order = Repository.Get(command.OrderId);
order.Approve(command.Comment, ServiceRequiredForOrderApproval);
Repository.Save(order);
}
}
Note that although my command handler (app service in a way) deals with repositories, my domain (order aggregate) is persistence ignorant. It doesn't know anything about repositories of UnitOfWorks.
When I do need to spin up a UnitOfWork I can compose it using a Chain Of Responsibility pattern:
public class WithUnitOfWorkHandler<T> : IHandler<T> {
private readonly IHandler<T> _innerHandler;
public WithUnitOfWorkHandler(IHandler<T> innerHandler) {
_innerHandler = innerHandler;
}
public void Handle(T command) {
using(var ouw = new UnitOfWork()) {
_innerHandler.Handle(command);
uow.Commit();
}
}
}
Now I can "chain" any of my command handlers by "decorating" it with WithUnitOfWorkHandler.
And some of the handlers may even touch more than one repository or aggregate. Still, aggregates don't know anything about persistence, unit of works, transactions, etc.
Persistence ignorance means: The business layer has no knowledge and no dependency whatsoever on the concrete persistence system that is used under the hood (e.g. MS SQL Server, Oracle, XML files, whatever).
Thus, exposing an interface that abstracts away the concrete type of the datastore can never violate this principle.
Persistence Ignorance is a guideline, it is almost impossible to reach with actual languages and technologies. The Repository pattern and the Unit Of Work abstract the persistence related stuff and "hide" the Data Access Layer to the business code, but it is more a trick (a clean one) than an absolute solution. The presence or the need for something (an interface, a base class, an attribute...) that says "heyyy, there is something in here we want to hide..." violates PI. But for the moment, there is no better solution.
I'm still learning about DDD and I have these two (probably simple) questions:
If a Factory creates new object/graph/aggregate instances, but also "reconstitutes" objects/graphs from the Repository, then:
(1) Does your service layer functions/jobs/tasks/unit-of-work call into the Factory or a behavioural method on the Entity instance or a DomainService function? I'm lost as to the call stack based on the responsibility of these components.
(2) Do Entity instances even have "behavioural methods" like above? For example does a Post have p.UpdatePost(string bodyText) or is that not a concern of the domain model and so the same should be achieved with the Repository? Or the service layer function, should it be calling the Repository in this case and the entity instance simply have behavioural methods that are specific to the domain and not persistence? But then, why does it sound like "updating a post" is a domain function when that's the user's goal?
You can see I'm all over the place. Please help.
(1) Does your service layer functions/jobs/tasks/unit-of-work call into the Factory or a behavioral method on the Entity instance or a DomainService function? I'm lost as to the call stack based on the responsibility of these components.
Usually - top level retrieves necessary aggregate root and calls a function on it. Sometimes top level retrieves multiple aggregate roots and pass them to domain service, but not often because domain service is a quite strong sign that there is unrecognized aggregate root. At the end - top level ensures aggregate root is persisted.
(2) Do Entity instances even have "behavioural methods" like above? For example does a Post have p.UpdatePost(string bodyText) or is that not a concern of the domain model and so the same should be achieved with the Repository? Or the service layer function, should it be calling the Repository in this case and the entity instance simply have behavioural methods that are specific to the domain and not persistence? But then, why does it sound like "updating a post" is a domain function when that's the user's goal?
Yes, they do. Domain model should be aware of it's state changes. And that's much more beneficial as it seems at first. Great thing about this is that You gain extensibility point. If client will walk week later to You and say that he wants system to check additional things when user updates post - instead of searching every line of post.bodyText="new value", You will be able to go straight to post.UpdatePost method and attach necessary things there.
On the other hand - CRUD is not mutually exclusive with domain driven design. E.g. - in my application, management of users and their roles is uninteresting enough that I'm not even trying to model it granularly. You need to recognize parts what matters in business Your application is describing and working with.
Keep in mind that domain driven design makes sense for complex applications only. Simple blog application doesn't need it.
(3) Am I wrong in assuming that a service layer (not Domain Services) should encapsulate how an interface interacts with the Domain Layer?
As I see it - application services are more for orchestrating infrastructure. If there is no infrastructure involved - then application service loses value:
Application services basically are just facades. And every facade is bad if complexity it adds overweights problems it solves.
Inside domain:
//aggregate root is persistence ignorant.
//it shouldn't reference repository directly
public class Customer{
public string Name {get; private set;}
public static Customer Register(string name){
return new Customer(name);
}
protected Customer(string name){
//here it's aware of state changes.
//aggregate root changes it's own state
//instead of having state changed from outside
//through public properties
this.Name=name;
}
}
//domain model contains abstraction of persistence
public interface ICustomerRepository{
void Save(Customer customer);
}
Outside of domain:
public class CustomerRepository:ICustomerRepository{
//here we actually save state of customer into database/cloud/xml/whatever
public void Save(Customer customer){
//note that we do not change state of customer, we just persist it here
_voodoo.StoreItSomehow(customer);
}
}
//asp.net mvc controller
public class CustomerController{
public CustomerController(ICustomerRepository repository){
if (repository==null)throw new ArgumentNullException();
_repository=repository;
}
public ActionResult Register(string name){
var customer=Customer.Register(name);
_repository.Save(customer);
}
}