3

I was just hoping someone might be able to answer a few questions I have regarding the proper design of DTOs and Domain Models. I'm currently working on a project which involves leveraging a SOAP API.

Since we're working with SOAP we have the added benefit of generating a number of objects (DTOs) using the provided WSDL file. I've also subsequently built a number of commands which are allowing us to better encapsulate the intent of each request and simplify our interface to the API.

Each command handler has an associated "command" object which stores all of the necessary data to complete the request successfully.

I actually have three questions regarding this approach:

1) Is it unusual to call command handlers directly, thereby replacing the Command Bus as our invoker. Obviously we would lose the ability (at least in these select instances) to pass our commands into a queuing application which would allow us to execute the command in a non-sequential order.

However, I have identified some cases where we would actually need to parse the API response data on runtime. Could this be construed as a potential issue in the design of our application?

2) We're currently performing validation on the request data before passing it into our immutable command objects. However, there may be some cases where we need parse data from the DB/CLI prior to initializing our commands.

As a result, I've considered placing our validation logic into the command objects themselves. If validation fails then a exception might be thrown. Does this seem like a reasonable approach or am I again violating some design principles?

3) This may sound foolish but what exactly is the difference between a Domain Object and a DTO. I've heard the command objects themselves being referred to as a DTO yet they appear to belong to the domain layer (unless I'm mistaken).

Could the command object be more accurately described as a DTO or a Domain Model?

Thanks again. I do appreciate the feedback.

1

Unfortunately, I am not sure what your first question really means, therefore am unable to answer that, but I can help you with the rest.

2) We're currently performing validation on the request data before passing it into our immutable command objects. However, there may be some cases where we need parse data from the DB/CLI prior to initializing our commands.

As a result, I've considered placing our validation logic into the command objects themselves. If validation fails then a exception might be thrown. Does this seem like a reasonable approach or am I again violating some design principles?

What you should ask yourself is why are you dealing with invalid objects in the first place so that they require validation after they are created?

There are two situations why this could happen:

  • you do not trust the variables which are used when an object is created
  • you do not trust the objects themselves, you do not believe, that what they are doing with their data is perfect

Both of these cases are bad, but the second one is much worse.

The first case can and does happen, when you use variables, which the user has control over. Ie. he can input anything he want. This is something you really have no control over, so you need to make sure to check all options before creating an object.

The second case happens when, even though you model a class, you do not trust the class does what it is supposed to do. Do you perhaps expose too much of its internal structure, giving the programmer too many posibilities on how to operate on the object, perhaps having simple getters and setters, instead of actual methods?

I said the second case is worse. The reason why it is worse is because you do not trust your own domain. And if you cannot trust it, then who?

3) This may sound foolish but what exactly is the difference between a Domain Object and a DTO. I've heard the command objects themselves being referred to as a DTO yet they appear to belong to the domain layer (unless I'm mistaken).

DTO is a very simple and stupid object which consists only from getters and setters and glues small chunks of seemingly related information together to make one larger structure, so the data requests may be reduced.

Martin Fowler provides a simple example of a DTO object.


Domain model is an object containing your business logic. It is an object which should have no methods starting with the prefix set, it should follow the Tell, Don't ask principle, etc. The domain is the core of your application, the domain objects make sure, everything works well, they throw domain exceptions on invalid operations to make sure, they are always in a valid state.

The Tell, Don't ask principle provides encapsulation. It turns this:

if (personObject.GetSex() == "Male") { ... }

into this:

if (personObject.IsMale()) { ... }

The programmer using a Person class no longer cares, how the Sex property is represented, if it is a string, an enum or a boolean. The only thing it cares about is the Person::IsMale : boolean method.

Domain model summary: It is an object containing data which are directly tied to it (no persistence logic is allowed, the object has to deal only with its own data) and encapsulates operations to make sure once created, it will never find itself in an invalid state.

0

I'm not very experienced in these terms exactly but I've done a lot of serialization with really complex data types with all kinds of weird dependencies, so here's my 2 cents on managing your POCOs(TLDR; keep them POCO).

I have identified some cases where we would actually need to parse the API response data on runtime. Could this be construed as a potential issue in the design of our application?

Yes, You should be done with processing all the data you get as soon as you can. Otherwise you may end up with a lot of spaghetti code in all the layers of your application doing de/serialization 'on-demand'

I've considered placing our validation logic into the command objects themselves. If validation fails then a exception might be thrown. Does this seem like a reasonable approach or am I again violating some design principles?

Yes, some validation may require data from other objects, some may require time, some results of expensive algorithms used for validation may be needed to accessed multiple times. You may need to do something if the validation fails that may include telling the web server that the validation failed. for that You either have to use some delegates or events or do it right there. both mean calling the the network library, not pretty and of course creating other DTOs in a DTO, all kinds of spaghetti there.

  • So the commands should be more accurately described as POCOs rather than DTOs/Domain Models then? – user2308097 Jul 5 '15 at 17:53
  • @user2308097 I guess so, but the principle should be the same. – Behrooz Jul 6 '15 at 1:43

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.