Performant and Explicit Domain Models

Monday, June 4th, 2007.

Some Technical Difficulties

Ayende and I had an email conversation that started with me asking what would happen if I added an Order to a Customer’s “Orders” collection, when that collection was lazy loaded. My question was whether the addition of an element would result in NHibernate hitting the database to fill that collection. His answer was a simple “yes”. In the case where a customer can have many (millions) of Orders, that’s just not a feasible solution. The technical solution was simple – just define the Orders collection on the Customer as “inverse=true”, and then to save a new Order, just write:

session.Save( new Order(myCustomer) );

Although it works, it’s not “DDD compliant” 🙂

In Ayende’s post Architecting for Performance he quoted a part of our email conversation. The conclusion I reached was that in order to design performant domain models, you need to know the kinds of data volumes you’re dealing with. It affects both internals and the API of the model – when can you assume cascade, and when not. It’s important to make these kinds of things explicit in the Domain Model’s API.

How do you make “transparent persistence” explicit?

The problem occurs around “transparent persistence”. If we were to assume that the Customer object added the Order object to its Orders collection, then we wouldn’t have to explicitly save orders it creates, so we would write service layer code like this:

using (IDBScope scope = this.DbServices.GetScope(TransactionOption.On))

{

	IOrderCreatingCustomer c = this.DbServices.Get<IOrderCreatingCustomer>(msg.CustomerId);

	c.CreateOrder(message.OrderAmount);
	scope.Complete();

}

On the other hand, if we designed our Domain Model around the million orders constraint, we would need to explicitly save the order, so we would write service layer code like this:

using (IDBScope scope = this.DbServices.GetScope(TransactionOption.On))

{

	IOrderCreatingCustomer c = this.DbServices.Get<IOrderCreatingCustomer>(msg.CustomerId);

	IOrder o = c.CreateOrder(message.OrderAmount);

	this.DbServices.Save(o);
	scope.Complete();

}

But the question remains, how do we communicate these guidelines to service layer developers from the Domain Model? There are a number of ways, but it’s important to decide on one and use it consistently. Performance and correctness require it.

Solution 1: Explicitness via Return Type

The first way is a little subtle, but you can do it with the return type of the “CreateOrder” method call. In the case where the Domain Model wishes to communicate that it handles transparent persistence by itself, have the method return “void”. Where the Domain Model wishes to communicate that it will not handle transparent persistence, have the method return the Order object created.

Another way to communicate the fact that an Order has been created that needs to be saved is with events. There are two sub-ways to do so:

Solution 2: Explicitness via Events on Domain Objects

The first is to just define the event on the customer object and have the service layer subscribe to it. It’s pretty clear that when the service layer receives a “OrderCreatedThatRequiresSaving” event, it should save the order passed in the event arguments.

The second realizes that the call to the customer object may come from some other domain object and that the service layer doesn’t necessarily know what can happen as the result of calling some method on the aggregate root. The change of state as the result of that method call may permeate the entire object graph. If each object in the graph raises its own events, its calling object will have to propagate that event to its parent – resulting in defining the same events in multiple places, and each object being aware of all things possible with its great-grandchild objects. That is clearly bad.

What [ThreadStatic] is for

So, the solution is to use thread-static events.

[Sidebar] Thread-static events are just static events defined on a static class, where each event has the ThreadStaticAttribute applied to it. This attribute is important for server-side scenarios where multiple threads will be running through the Domain Model at the same time. The easiest thread-safe way to use static data is to apply the ThreadStaticAttribute.

Solution 3: Explicitness via Static Events

Each object raises the appropriate static event according to its logic. In our example, Customer would call:

DomainModelEvents.RaiseOrderCreatedThatRequiresSavingEvent(newOrder);

And the service layer would write:

DomainModelEvents.OrderCreatedThatRequiresSaving +=

	delegate(object sender, OrderEventArgs e) { this.DbServices.Save(e.Order); };

The advantage of this solution is that it requires minimal knowledge of the Domain Model for the Service Layer to correctly work with it. It also communicates that anything that doesn’t raise an event will be persisted transparently behind the appropriate root object.

Statics and Testability

I know that many of you are wondering if I am really advocating the use of statics. The problem with most static classes is that they hurt testability because they are difficult to mock out. Often statics are used as Facades to hide some technological implementation detail. In this case, the static class is an inherent part of the Domain Model and does not serve as a Facade for anything.

When it comes to testing the Domain Model, we don’t have to mock anything out since the Domain Model is independent of all other concerns. This leaves us with unit testing at the single Domain Class level, which is pretty useless unless we’re TDD-ing the design of the Domain Model, in which case we’ll still be fiddling around with a bunch of classes at a time. Domain Models are best tested using State-Based Testing; get the objects into a given state, call a method on one of them, assert the resulting state. The static events don’t impede that kind of testing at all.

What if we used Injection instead of Statics?

Also, you’ll find that each Service Layer class will need to subscribe to all the Domain Model’s events, something that is easily handled by a base class. I will state that I have tried doing this without a static class, and injecting that singleton object into the Service Layer classes, and in that setter having them subscribe to its events. This was also pulled into a base class. The main difference was that the Dependency Injection solution required injecting that object into Domain Objects as well. Personally, I’m against injection for domain objects. So all in all, the static solution comes with less overhead than that based on injection.

Summary

In summary, beyond the “technical basics” of being aware of your data volumes and designing your Domain Model to handle each use case performantly, I’ve found these techniques useful for designing its API as well as communicating my intent around persistence transparency. So give it a try. I’d be grateful to hear your thoughts on the matter as well as what else you’ve found that works.

Fetching Strategy Design – showing how to separate the concern of eager loading from both your Domain Model and your Service Layer.
Better Domain-Driven Design Implementation – showing the basics of how valuable interfaces between your Domain Model and the Service Layer can be.
Lazy Loading, and how messaging fixes everything again – describing the advantage of O/R mapping your message classes as well.
Query Objects vs Methods on a Repository – discussing the scalability (in terms of number of developers and queries) of Query Objects.

If you liked this article, you might also like articles in these categories:

If you've got a minute, you might enjoy taking a look at some of my best articles.
I've gone through the hundreds of articles I've written over the past 6 years and put together a list of the best ones as ranked by my 5000+ readers.
You won't be disappointed.

If you'd like to get new articles sent to you when they're published, it's easy and free.
Subscribe right here.

Follow me on Twitter @UdiDahan.

Something on your mind? Got a question? I'd be thrilled to hear it.
Leave a comment below or email me, whatever works for you.

thesoftwaresimplist Says:
June 4th, 2007 at 10:38 pm
It appears that NHibernate Bags don’t have this drawback:

http://www.hibernate.org/hib_docs/nhibernate/1.2/reference/en/html/performance.html#performance-collections-mostefficentinverse

Frederik Says:
June 17th, 2007 at 10:03 am
Another thing one might ask is, why should your Customer Entity contain a collection of Order objects ?
This means that, whenever you do something with your Customer, you drag all his orders (maybe 1000) with you. In which case(s) would it be necessary to have the Orders of a Customer ? In which cases do you want to have an overview of all the Orders of a particular Customer ?
I always ask myself this question in such scenario’s, and very often, I decide to not have a collection of (in this example) Orders in my Customer entity. Instead, I opt for a method in my OrderRepository which is called ‘GetAllOrdersForCustomer( Customer c )’. Then, in the cases where I do need the Orders of a particular Customer, I just have to call this method …
In my opinion -and in this example- it is far more usefull to know the Customer to which an Order belongs, then to having a collection of Orders in the customer entity.

Colin Jack Says:
June 20th, 2007 at 6:09 am
Interesting topic. Personally I don’t like the event idea, you only really want there to be one subscriber and they absolutely should be subscribing so events don’t seem to me to be appropriate.

Equally I don’t like being forced to use relationships from the one end (e.g. Order to Client) just for performance. Designing your associations like that is going to result in a mess in some cases, it might work for Order/Customer but in complex cases where the classes are in seperate modules it can be messy. I’ve posted about this sort of issue here:

http://colinjack.blogspot.com/2007/03/persistence-ignorance-and-associations.html

One way I handle this is to have these sorts of calls go through a Coordination layer that will ensue the new Orders are saved:

http://colinjack.blogspot.com/2007/02/domain-coordination-layer.html

Having said all that we use bags with NHibernate so it is not really an issue for us.

Any thoughts?

thesoftwaresimplist Says:
June 20th, 2007 at 8:57 am
Frederik,

> Another thing one might ask is, why should your Customer Entity contain a collection of Order objects ? In which case(s) would it be necessary to have the Orders of a Customer ?

We may have business rules that impact whether the customer is preferred (and thus gets a discount) or not based on some calculation on their order history.

> This means that, whenever you do something with your Customer, you drag all his orders (maybe 1000) with you.

That was my point – I don’t want to do that. My problem was that even if the collection of orders was lazy loaded, adding a new order would cause all those orders to be loaded. This turns out to be a non-problem with NHibernate “bags” which don’t load when an object is added.

> I decide to not have a collection of (in this example) Orders in my Customer entity. Instead, I opt for a method in my OrderRepository which is called ‘GetAllOrdersForCustomer( Customer c )’. Then, in the cases where I do need the Orders of a particular Customer, I just have to call this method.

Since I implement my business rules within the Domain Model, the problem I have with that is that I don’t want my Domain Objects calling into Repositories. That’s not to say that such a query would not be valuable on its own.

Does that answer your questions?

thesoftwaresimplist Says:
June 20th, 2007 at 10:45 am
Colin,

> Interesting topic.

Glad you liked it.

> Personally I don’t like the event idea, you only really want there to be one subscriber and they absolutely should be subscribing so events don’t seem to me to be appropriate.

Events / delegates are used to keep the logic about when things need to occur in the Domain Model, while keeping it indpendent of how the wiring occurs. It’s all about dependency management, but you’re entitled to your own taste.

> Equally I don’t like being forced to use relationships from the one end (e.g. Order to Client) just for performance.

Me neither. That’s why I make use of interfaces for hiding these sorts of implementation details.

> Designing your associations like that is going to result in a mess in some cases, it might work for Order/Customer but in complex cases where the classes are in seperate modules it can be messy.

I’ve done the multi-module thing and it’s usually doable by using IList of IEntity for holding cross module collections.

> One way I handle this is to have these sorts of calls go through a Coordination layer that will ensue the new Orders are saved.

Personally, I prefer events to coupling the Domain Model to some other Coordination Layer, even if that’s behind some interface, it forces me to inject that dependency into my Domain Objects. That kind of complexity I don’t want.

I think that we can agree to disagree on this one 🙂

Frederik Says:
June 20th, 2007 at 1:47 pm
Hmm, why don’t you want to call Repositories from your Domain Entities ? As I read in Evans’ book: repositories are first class citizens of the domain layer; they are part of the domain.

I understand that you might have some business logic which decides wether a customer is a gold customer or not, based on the orders that this customer has made, but then, it might be necessary that you iteratoe over all the Orders in the OrderCollection of that Customer. When this customer has a lot of Orders, this can put a weight on performance. Wouldn’t it then be better to use an injected OrderRepository to get the Orders you’re really interested in to perform this logic out of the persistence store ?

thesoftwaresimplist Says:
June 20th, 2007 at 2:28 pm
Frederik,

> Hmm, why don’t you want to call Repositories from your Domain Entities?

Separation of Concerns – Domain Objects deal with business logic only and know nothing about persistence, and thus are highly testable while Repositories are all about persistence.

> Wouldn’t it then be better to use an injected OrderRepository to get the Orders you’re really interested in to perform this logic out of the persistence store ?

I actually “denormalize” both my Domain Objects and database to handle these kinds of performance requirements. By having the running total of all orders (if that’s what the rules state) as a persistent member on my Customer class, and make sure that value is consistent, I won’t need to iterate. Injecting an OrderRepository wouldn’t solve the performance issue.

Does that make sense?

Frederik Says:
June 22nd, 2007 at 4:04 am
Hmm, well, then we have a different view on repositories. In my opinion, they’re part of the Domain and as such, they can contain specialized methods that are required to do some Business Logic. (For Instance ‘GetOrderTotalForCustomerInPeriod’.
For me, a repository is more then just a way of retrieving / storing data.

About denormalizing: sometimes this is indeed necessary; however, I do not like to ‘sacrifice’ my DB model just because it is necessary if I want to keep all my logic in the Domain Entities.
Your domain layer might make sure that everything is kept consistent, but it is maybe also possible that other (legacy) applications use the same database. Then, the consistency is not guaranteed.

thesoftwaresimplist Says:
June 22nd, 2007 at 7:58 am
Frederik,

I prefer handling ‘GetOrderTotalForCustomerInPeriod’ using the Query Object Pattern, as I wrote about here:

http://udidahan.weblogs.us/2007/03/28/query-objects-vs-methods-on-a-repository/

The necessity comes out of performance. Keeping as much of my logic as possible in the Domain Model has been important for me in terms of testability.

I agree that the issue of using the database as the integration platform creates problems that we may not be able to do away with in one shot. However, I try not to propogate this anti-pattern and phase things out over time. As long as we’re going in the right direction, I’m willing to make some sacrifices on correctness in the short term.

I’m going to be putting an article out in the next month or so on the use of DBs as integration platforms and how that’s changing with SOA. I’d be interested in hearing more about your experiences in that area.

Colin Jack Says:
November 16th, 2007 at 1:55 pm
Just re-read this post and got this bit:

“When it comes to testing the Domain Model, we don’t have to mock anything out since the Domain Model is independent of all other concerns.”

I agree with you but wondered how you go about it, do you use any mocking at all? Do you test at the aggregate level? What do you do about testing encapsulated classes (e.g. state objects or strategies), do you just test them through the effects they have on the public methods?

Just interested as its a topic thats not discussed enough.

> Personally, I prefer events to coupling the Domain Model to some
> other Coordination Layer, even if that’s behind some interface, it > forces me to inject that dependency into my Domain Objects. That
> kind of complexity I don’t want.

I should also say, on the coordination layer front its the coordination layer that calls the domain classes not the other way round. The domain classes only know about other domain classes, not about any layers above or at the same level.

thesoftwaresimplist Says:
November 16th, 2007 at 5:22 pm
Colin,

For testing the correctness of the Domain Model (unit testing), no mocking is done. I test according to interfaces, which may or may not be at the aggregate level. For instance, Orders belong to a Customer, the aggregate – I do test Orders directly.

About encapsulated classes, I test through the external interface – focusing on externally visible effects.

About coordination, Domain Objects can call back into Coordination Layer classes – without depending on them, using events. Does that make sense?

Nuno Lopes Says:
February 6th, 2009 at 7:28 am
Again late to the party so sorry to bring this up again as I think it is crucial.

Although Bags resolve the problem at hand, and sometimes are necessary, prior to it we have another problem (actually I have been seeing this problem in a lot of examples).

Again aggregates. We have found more then handful types of aggregations depending on the type of collaboration (Composite-item, Container-item, Group-Member, Party-Transaction, Template-Item, etc etc etc etc etc).

The collaboration pattern of choice have specific rules about who directs what in a collaboration. This in contrast of one recepee fits all for any kind of aggregation (just put add remove in the aggregator and it rules everything about creating its “items”). The problem is that say the object being aggregated (The Order) can also be agregated by some other object at the time of creation (Place?).

For instance:

“A Customer making an Order of products on a Place”.

So who does create the Order? The customer (or its Role) that makes it, or the place (or its Role) that accepts it?

I know that everyone will say that is the Customer that makes it (good OO). But this question makes sense in a lot contexts (too many).

I’ve found that in situations such as this it is preferable to do something like this on the service:

Order aOrder = new Order(aCustomer, aPlace);
// add products.
tx.Commit.

Actually I tend to do this everytime except for aggregations that imply ownership … aka composition (which is different for assembly-part).

An example of this we have right here. The Order and its OrderLines. An Order is Composed by OrderLines. Which is not the case of Customer and Order (A customer is not composed by its orders). In most business cases an Order (Transaction) can last longer then then any info put in the Customer entity as it can change (Name, Delivery Address, Payment etc etc etc should be all be within the Order that can refer the customer for completness but it is not essential).

In my experience, I’ve found also that in most Composite collaborations actually the number of items in the composite (Say Building-Floor-Room, or Organization-Department-Subdepartement) is quite small or hierarchical.

In other words, except for the composites objects usually aggregators don’t create it’s “items”. That is easy to explain to any team member coding the service layer.

So the problem is solved at the Domain Model and OOP practice rather then technology and it becomes natural. Furthermore the technical solution given by Ayende works quite well.

Notice that I’m not proposing that the Customer shouldn’t be able have and provide a list of its Orders, it should. Nevertheless if it is “him” solely governs the list to the point it creates the items in it is another matter.

Does this make sense?

Nuno

Nuno Lopes Says:
February 6th, 2009 at 7:39 am
By the way I liked the article about Domain Fluence.

One might argue that this approach is not Domain Fluent, but I can equally argue that it is.

Nuno

Nuno Lopes Says:
February 6th, 2009 at 10:57 am
For completeness of my reasoning above.

There might be a business rule that there “must” be no Customer without at least one order. Ideally we would do:

Customer aCustomer = new Customer(aOrder);

But the approach I gave required that the Customer to be “born” first.

This problem reminds me a riddle: “Who was born first, the chicken or the Egg?”

So how do we validate this? Well we use Validators much the explained in some powerpoint of Udi (where an event is trapped right before the object is to be stored).

Validators very helpfull for invariants. That is rules that once true are required to be true as long as the object exists in the system.

Nuno