Nov 6, 2008

Notes for Domain Driven Design Quickly

What is domain driven design?

The best way to do it is to make software a reflection of the domain. Software needs to incorporate the core concepts and elements of the domain, and to precisely realize the relations between them. Sofeware has to model the domain.

Somebody without knowledge of banking should be able to learn a lot just buy reading the code in a domain model. This is essential. Software which does not have its roots planted deep into the the domain will not react well to change over time.

We need to create an abstraction of the domain. It is a domain model. It is not a particular diagram, it is the idea that the diagram is intended to convey. It is not just the knowledge in a domain expert's head, It is a rigorously organized and selective abstraction of that knowledge. It can be in the form a diagram , carefully written code or just english sentence.

A specific domain could be more than a human can handle at one time. We need to orgainze information, to systematize, to divid it up in smaller pieces, to group those pieces into logical modules, and take one at a time and deal with it.

A model is the essence of the software, but need to create ways to express it, to communicate it with others. We need to cummunicate the model. There are different ways to do that. One is graphical: diagrams, use cases, drawings, pictures, etc. Another is writing. We write down our vision about the domain. Another is language. We can and we should create a language to communicate specific issues about the domain. The right way to design software is domain driven design. It combines design and development practice, and shows how design and development can work together to create a better solution. Good design will accelerate the development, while feed back coming from the development process will enhance the design.

You and the domain experts need to exchanging knowledge. You start asking questions, and they respond. While do do that, they dig essential concepts out of the domain. Those concepts may come out unpolished and disorganized, but nonetheless they are essential for understanding the domain. You need to learn as much as possible about the domain from the experts. And by putting the right questions, and processing the information in the right way, you and the experts will start to sketch a view of the domain, a domain model. This view is neither complete nor correct, but it is the start you need. Try to figure out the essential concepts of the domain. This is an important part of the design. Usually there are long discussions between software architects or developers and the domain experts. The software specialists want to extract knowledge from the domain experts, and they also have to transform it into a useful form. At some point, they might want to create an early prototype to see how it works so far. While doing that they may find some issues with their model, or their approach, and may want to change the model. The communication is not only one way, from the domain experts to the software architect and further to the developers. There is also feedback, which helps create a better model, and a clearer and more correct understanding of the domain. Domain experts know their area of expertise well, but they organize and use their knowledge in a specific way, which is not always the best to be implemented into a software system. The analytical mind of the software designer helps unearth some of the key concepts of the domain during discussions with domain experts, and also help construct a structure for future discussions as we will see in the next chapter. We, the software specialists (software architects and developers) and the domain experts, are creating the model of the domain together, and the model is the place where those two areas of expertise meet. This might seem like a very time consuming process, and it is, but this is how it should be, because in the end the software’s purpose is to solve business problems in a real life domain, so it has to blend perfectly with the domain.

Model Driven Design

After we create a great model, we need to implement the model in code, propertyly transfer it into code. Any domain can be expressed with many models, and any model can be expressed in various ways in code. For each particular problem there can be more than one solution. Which one do we choose? Having one analytically correct model does not mean the model can be directly expressed in code. Or maybe its implementation will break some software design principles, which is not advisable. It is important to choose a model which can be easily and accurately put into code. The basic question here is: how do we approach the transition from model to code?

The disconnection between model and code design

It happens that software analysts work with business domain experts for months, discover the fundamental elements of the domain, emphasize the relationshipbs between them, and create a correct model, which accurately captures the domain. Then the model is passed on to the software developers. The developers might look at the model and discover that some of the concepts or relationships found in it cannot be properly expressed in code. So they use the model as the original source of inspiration, but they create their own design which borrows some of the ideas from the model, and adds some of their own. The development process continues further, and more classes are added to the code, expanding the divide between the original model and the final implementation. The good end result is not assured. Good developers might pull together a product which works, but will it stand the trials of time? Will it be easily extendable? Will it be easily maintainable?

Any domain can be expressed with many models, and any model can be expressed in various ways in code. For each particular problem there can be more than one solution. Which one do we choose? Having one analytically correct model does not mean the model can be directly expressed in code. Or maybe its implementation will break some software design principles, which is not advisable. It is important to choose a model which can be easily and accurately put into code. The basic question here is: how do we approach the transition from model to code?

One of the recommended design techniques is the so called analysis model, which is seen as separate from code design and is usually done by different people. The analysis model is the result of business domain analysis, resulting in a model which has no consideration for the software used for implementation. Such a model is used to understand the domain. A certain level of knowledge is built, and the model resulting may be analytically correct. Software is not taken into account at this stage because it is considered to be a confusing factor. This model reaches the developers which are supposed to do the design. Since the model was not built with design principles in mind, it probably won’t serve that purpose well. The developers will have to adapt it, or to create a separate design. And there is no longer a mapping between the model and the code. The result is that analysis models are soon abandoned after coding starts.

One of the main issues with this approach is that analysts cannot foresee some of the defects in their model, and all the intricacies of the domain. The analysts may have gone into too much detail with some of the components of the model, and have not detailed enough others. Very important details are discovered during the design and implementation process. A model that is truthful to the domain could turn out to have serious problems with object persistence, or unacceptable performance behavior. Developers will be forced to make some decisions on their own, and will make design changes in order to solve a real problem which was not considered when the model was created. They create a design that slips away from the model, making it less relevant.

If the analysts work independently, they will eventually create a model. When this model is passed to the designers, some of the analysts’ knowledge about the domain and the model is lost. While the model might be expressed in diagrams and writing, chances are the designers won’t grasp the entire meaning of the model, or the relationships between some objects, or their behavior. There are details in a model which are not easily expressed in a diagram, and may not be fully presented even in writing. The developers will have a hard time figuring them out. In some cases they will make some assumptions about the intended behavior, and it is possible for them to make the wrong ones, resulting in incorrect functioning of the program.

Analysts have their own closed meetings where many things are discussed about the domain, and there is a lot of knowledge sharing. They create a model which is supposed to contain all that information in a condensed form, and the developers have to assimilate all of it by reading the documents given to them. It would be much more productive if the developers could join the analyst meetings and have thus attain a clear and complete view of the domain and the model before they start designing the code.

Connect model and code design

A better approach is to closely relate domain modeling and design. The model should be constructed with an eye open to the software and design considerations. Developers should be included in the modeling process. The main idea is to choose a model which can be appropriately expressed in software, so that the design process is straightforward and based on the model. Tightly relating the code to an underlying model gives the code meaning and makes the model relevant.

Getting the developers involved provides feedback. It makes sure that the model can be implemented in software. If something is wrong, it is identified at an early stage, and the problem can be easily corrected.

Those who write the code should know the model very well, and should feel responsible for its integrity. They should realize that a change to the code implies a change to the model; otherwise they will refactor the code to the point where it no longer expresses the original model. If the analyst is separated from the implementation process, he will soon lose his concern about the limitations introduced by development. The result is a model which is not practical.

Any technical person contributing to the model must spend some time touching the code, whatever primary role he or she plays on the project. Anyone responsible for changing code must learn to express a model through the code. Every developer must be involved in some level of discussion about the model and have contact with domain experts. Those who contribute in different ways must consciously engage those who touch the code in a dynamic exchange of model ideas through the Ubiquitous Language.

If the design, or some central part of it, does not map to the domain model, that model is of little value, and the correctness of the software is suspect. At the same time, complex mappings between models and design functions are difficult to understand and, in practice, impossible to maintain as the design changes. A deadly divide opens between analysis and design so that insight gained in each of those activities does not feed into the other.

Design a portion of the software system to reflect the domain model in a very literal way, so that mapping is obvious. Revisit the model and modify it to be implemented more naturally software, even as you seek to make it reflect deeper insight in the domain. Demand a single model that serves both purpose well, in addition to supporting a fluent Ubiquitous Language.

Draw from the model the terminology used in the design and the basic assignment of responsibilities. The code becomes an expression of the model, so a change to the code may be a change to the model. Its effect must ripple through the rest of the project’s activities accordingly.

To tightly tie the implementation to a model usually requires software development tools and languages that support a modeling paradigm, such as object-oriented programming.

The building blocks of a model driven design

When we create a software application, a large part of the application is not directly related to the domain, but it is part of the infrastructure or serves the software itself. It is possible and ok for the domain part of an application to be quite small compared to the rest, since a typical application contains a lot of code related to database access, file or network access, user interfaces, etc.

In an object-oriented program, UI, database, and other support code often gets written directly into the business objects. Additional business logic is embedded in the behavior of UI widgets and database scripts. This some times happens because it is the easiest way to make things work quickly.

However, when domain-related code is mixed with the other layers, it becomes extremely difficult to see and think about. Superficial changes to the UI can actually change business logic. To change a business rule may require meticulous tracing of UI code, database code, or other program elements. Implementing coherent, model-driven objects becomes impractical. Automated testing is awkward. With all the technologies and logic involved in each activity, a program must be kept very simple or it becomes impossible to understand.

Therefore, partition a complex program into LAYERS. Develop a design within each LAYER that is cohesive and that depends only on the layers below. Follow standard architectural patterns to provide loose coupling to the layers above. Concentrate all the code related to the domain model in one layer and isolate it from the user interface, application, and infrastructure code. The domain objects, free of the responsibility of displaying themselves, storing themselves, managing application tasks, and so forth, can be focused on expressing the domain model. This allows a model to evolve to be rich enough and clear enough to capture essential business knowledge and put it to work.

  • User Interface (Presentation Layer)

    Responsible for presenting information to the user and interpreting user commands.

  • Application Layer

    This is a thin layer which coordinates the application activity. It does not contain business logic. It does not hold the state of the business objects, but it can hold the state of an application task progress.

  • Domain Layer

    This layer contains information about the domain. This is the heart of the business software. The state of business objects is held here. Persistence of the business objects and possibly their state is delegated to the infrastructure layer.

  • Infrastructure Layer

    This layer acts as a supporting library for all the other layers. It provides communication between layers, implements persistence for business objects, contains supporting libraries for the user interface layer, etc.

It is important to divide an application in separate layers, and establish rules of interactions between the layers. If the code is not clearly separated into layers, it will soon become so entangled that it becomes very difficult to manage changes. One simple change in one section of the code may have unexpected and undesirable results in other sections. The domain layer should be focused on core domain issues. It should not be involved in infrastructure activities. The UI should neither be tightly connected to the business logic, nor to the tasks which normally belong to the infrastructure layer. An application layer is necessary in many cases. There has to be a manager over the business logic which supervises and coordinates the overall activity of the application.

Entity

There is a category of objects which seem to have an identity, which remains the same throughout the states of the software. For these objects it is not the attributes which matter, but a thread of continuity and identity, which spans the life of a system and can extend beyond it. Such objects are called Entities

Therefore, implementing entities in software means creating identity. For a person it can be a combination of attributes: name, date of birth, place of birth, name of parents, current address. The Social Security number is also used in US to create identity. For a bank account the account number seems to be enough for its identity. Usually the identity is either an attribute of the object, a combination of attributes, an attribute specially created to preserve and express identity, or even a behavior. It is important for two objects with different identities to be to be easily distinguished by the system, and two objects with the same identity to be considered the same by the system. If that condition is not met, then the entire system can become corrupted.

Value Objects

We may be tempted to make all objects entities. Entities can be tracked. But tracking and creating identity comes with a cost. We need to make sure that each instance has its unique identity, and tracking identity is not very simple

If we don't need to keep track of an object, then it is value objects.

A value object can be immutable or mutable. But it is highly recommended that value objects be immutable. They are created with a constructor, and never modified during their life time.When you want a different value for the object, you simply create another one. This has important consequences for the design. Being immutable, and having no identity, Value Objects can be shared. That can be imperative for some designs. Immutable objects are sharable with important performance implications. They also manifest integrity, i.e. data integrity. Imagine what it would mean to share an object which is not immutable. An air travel booking system could create objects for each flight. One of the attributes could be the flight code. One client books a flight for a certain destination. Another client wants to book the same flight. The system chooses to reuse the object which holds the flight code, because it is about the same flight. In the meantime, the client changes his mind, and chooses to take a different flight. The system changes the flight code because this is not immutable. The result is that the flight code of the first client changes too.

One golden rule is: if Value Objects are shareable, they should be immutable. Value Objects should be kept thin and simple. When a Value Object is needed by another party, it can be simply passed by value, or a copy of it can be created and given. Making a copy of a Value Object is simple, and usually without any consequences. If there is no identity, you can make as many copies as you wish, and destroy all of them when necessary.

Services

A Service should not replace the operation which normally belongs on domain objects. We should not create a Service for every operation needed. But when such an operation stands out as an important concept in the domain, a Service should be created for it. There are three characteristics of a Service:

  1. The operation performed by the Service refers to a domain concept which does not naturally belong to an Entity or Value Object.
  2. The operation performed refers to other objects in the domain.
  3. The operation is stateless.

When a significant process or transformation in the domain is not a natural responsibility of an Entity or Value Object, add an operation to the model as a standalone interface declared as a Service. Define the interface in terms of the language of the model and make sure the operation name is part of the Ubiquitous Language. Make the Service stateless.

While using Services, is important to keep the domain layer isolated. It is easy to get confused between services which belong to the domain layer, and those belonging to the infrastructure. There can also be services in the application layer which adds a supplementary level of complexity. Those services are even more difficult to separate from their counterparts residing in the domain layer. While working on the model and during the design phase, we need to make sure that the domain level remains isolated from the other levels.

Both application and domain Services are usually built on top of domain Entities and Values providing required functionality directly related to those objects. Deciding the layer a Service belongs to is difficult. If the operation performed conceptually belongs to the application layer, then the Service should be placed there. If the operation is about domain objects, and is strictly related to the domain, serving a domain need, then it should belong to the domain layer.