Domain-Driven Design

by Eric Evans

  assertTrue(a0.assignedRules().contains(minWidth4));

  assertEquals(minWidth4, a0.getRule(MIN_WIDTH));

  assertEquals(minWidth4, a1.getRule(MIN_WIDTH));

  }

  An interactive user interface could present a list of buses, allowing the user to assign rules to each, or it could read from a file of rules for backward compatibility. A façade makes access simple for either interface. Its implementation echoes the test:

  public void assignBusRule(String busName, String ruleType,

  double parameter){

  Bus bus = BusRepository.getByName(busName);

  bus.assignRule(NetRule.create(ruleType, parameter));

  }

  Finishing:

  NetRuleExport.write(aFileName, NetRepository.allNets());

  (The service asks each Net for assignedRules(), and then writes them fully expanded.)

  Of course, if there were only one operation (as in the example), the script-based approach might be just as practical. But in reality, there were 20 or more. The MODEL-DRIVEN DESIGN scales easily and can include constraints on combining rules and other enhancements.

  The second design also accommodates testing. Its components have well-defined interfaces that can be unit-tested. The only way to test the script is to do an end-to-end file-in/file-out comparison.

  Keep in mind that such a design does not emerge in a single step. It would take several iterations of refactoring and knowledge crunching to distill the important concepts of the domain into a simple, incisive model.

  Letting the Bones Show: Why Models Matter to Users

  In theory, perhaps, you could present a user with any view of a system, regardless of what lies beneath. But in practice, a mismatch causes confusion at best—bugs at worst. Consider a very simple example of how users are misled by superimposed models of bookmarks for Web sites in current releases of Microsoft Internet Explorer.1

  A user of Internet Explorer thinks of “Favorites” as a list of names of Web sites that persist from session to session. But the implementation treats a Favorite as a file containing a URL, and whose filename is put in the Favorites list. That’s a problem if the Web page title contains characters that are illegal in Windows filenames. Suppose a user tries to store a Favorite and types the following name for it: “Laziness: The Secret to Happiness”. An error message will say: “A filename cannot contain any of the following characters: / : * ? " < > | ”. What filename? On the other hand, if the Web page title already contains an illegal character, Internet Explorer will just quietly strip it out. The loss of data may be benign in this case, but not what the user would have expected. Quietly changing data is completely unacceptable in most applications.

  MODEL-DRIVEN DESIGN calls for working with only one model (within any single context, as will be discussed in Chapter 14). Most of the advice and examples go to the problems of having separate analysis models and design models, but here we have a problem arising from a different pair of models: the user model and the design/implementation model.

  Of course, an unadorned view of the domain model would definitely not be convenient for the user in most cases. But trying to create in the UI an illusion of a model other than the domain model will cause confusion unless the illusion is perfect. If Web Favorites are actually just a collection of shortcut files, then expose this fact to the user and eliminate the confusing alternative model. Not only will the feature be less confusing, but the user can then leverage what he knows about the file system to deal with Web Favorites. He can reorganize them with the File Explorer, for example, rather than use awkward tools built into the Web browser. Informed users would be more able to exploit the flexibility of storing Web shortcuts anywhere in the file system. Just by removing the misleading extra model, the power of the application would increase and become clearer. Why make the user learn a new model when the programmers felt the old model was good enough?

  Alternatively, store the Favorites in a different way, say in a data file, so that they can be subject to their own rules. Those rules would presumably be the naming rules that apply to Web pages. That would again provide a single model. This one tells the user that everything he knows about naming Web sites applies to Favorites.

  When a design is based on a model that reflects the basic concerns of the users and domain experts, the bones of the design can be revealed to the user to a greater extent than with other design approaches. Revealing the model gives the user more access to the potential of the software and yields consistent, predictable behavior.

  Hands-On Modelers

  Manufacturing is a popular metaphor for software development. One inference from this metaphor: highly skilled engineers design; less skilled laborers assemble the products. This metaphor has messed up a lot of projects for one simple reason—software development is all design. All teams have specialized roles for members, but overseparation of responsibility for analysis, modeling, design, and programming interferes with MODEL-DRIVEN DESIGN.

  On one project, my job was to coordinate different application teams and help develop the domain model that would drive the design. But the management thought that modelers should be modeling, and that coding was a waste of those skills, so I was in effect forbidden to program or work on details with programmers.

  Things seemed to be OK for a while. Working with domain experts and the development leads of the different teams, we crunched knowledge and refined a nice core model. But that model was never put to work, for two reasons.

  First, some of the model’s intent was lost in the handoff. The overall effect of a model can be very sensitive to details (as will be discussed in Parts II and III), and those details don’t always come across in a UML diagram or a general discussion. If I could have rolled up my sleeves and worked with the other developers directly, providing some code to follow as examples, and providing some close support, the team could have taken up the abstractions of the model and run with them.

  The other problem was the indirectness of feedback from the interaction of the model with the implementation and the technology. For example, certain aspects of the model turned out to be wildly inefficient on our technology platform, but the full implications didn’t trickle back to me for months. Relatively minor changes could have fixed the problem, but by then it didn’t matter. The developers were well on their way to writing software that did work—without the model, which had been reduced to a mere data structure, wherever it was still used at all. The developers had thrown the baby out with the bathwater, but what choice did they have? They could no longer risk being saddled with the dictates of the architect in the ivory tower.

  The initial circumstances of this project were about as favorable to a hands-off modeler as they ever are. I already had extensive hands-on experience with most of the technology used on the project. I had even led a small development team on the same project before my role changed, so I was familiar with the project’s development process and programming environment. Even those factors were not enough to make me effective, given the separation of modeler from implementation.

  If the people who write the code do not feel responsible for the model, or don’t understand how to make the model work for an application, then the model has nothing to do with the software. If developers don’t realize that changing code changes the model, then their refactoring will weaken the model rather than strengthen it. Meanwhile, when a modeler is separated from the implementation process, he or she never acquires, or quickly loses, a feel for the constraints of implementation. The basic constraint of MODEL-DRIVEN DESIGN—that the model supports an effective implementation and abstracts key domain knowledge—is half-gone, and the resulting models will be impractical. Finally, the knowledge and skills of experienced designers won’t be transferred to other developers if the division of labor prevents the kind of collaboration that conveys the subtleties of coding a MODEL-DRIVEN DESIGN.

  The need for HANDS-ON MODELERS does not mean that team members cannot have specialize
d roles. Every Agile process, including Extreme Programming, defines roles for team members, and other informal specializations tend to emerge naturally. The problem arises from separating two tasks that are coupled in a MODEL-DRIVEN DESIGN, modeling and implementation.

  The effectiveness of an overall design is very sensitive to the quality and consistency of fine-grained design and implementation decisions. With a MODEL-DRIVEN DESIGN, a portion of the code is an expression of the model; changing that code changes the model. Programmers are modelers, whether anyone likes it or not. So it is better to set up the project so that the programmers do good modeling work.

  Therefore:

  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.

  The sharp separation of modeling and programming doesn’t work, yet large projects still need technical leaders who coordinate high-level design and modeling and help work out the most difficult or most critical decisions. Part IV, “Strategic Design,” deals with such decisions and should stimulate ideas for more productive ways to define the roles and responsibilities of high-level technical people.

  Domain-driven design puts a model to work to solve problems for an application. Through knowledge crunching, a team distills a torrent of chaotic information into a practical model. A MODEL-DRIVEN DESIGN intimately connects the model and the implementation. The UBIQUITOUS LANGUAGE is the channel for all that information to flow between developers, domain experts, and the software.

  The result is software that provides rich functionality based on a fundamental understanding of the core domain.

  As mentioned, success with MODEL-DRIVEN DESIGN is sensitive to detailed design decisions, which is the subject of the next several chapters.

  II: The Building Blocks of a Model-Driven Design

  To keep a software implementation crisp and in lockstep with a model, in spite of messy realities, you must apply the best practices of modeling and design. This book is not an introduction to object-oriented design, nor does it propose radical design fundamentals. Domain-driven design shifts the emphasis of certain conventional ideas.

  Certain kinds of decisions keep the model and implementation aligned with each other, each reinforcing the other’s effectiveness. This alignment requires attention to the details of individual elements. Careful crafting at this small scale gives developers a steady platform from which to apply the modeling approaches of Parts III and IV.

  The design style in this book largely follows the principle of “responsibility-driven design,” put forward in Wirfs-Brock et al. 1990 and updated in Wirfs-Brock 2003. It also draws heavily (especially in Part III) on the ideas of “design by contract” described in Meyer 1988. It is consistent with the general background of other widely held best practices of object-oriented design, which are described in such books as Larman 1998.

  As a project hits bumps, large or small, developers may find themselves in situations that make those principles seem inapplicable. To make the domain-driven design process resilient, developers need to understand how the well-known fundamentals support MODEL-DRIVEN DESIGN, so they can compromise without derailing.

  The material in the following three chapters is organized as a “pattern language” (see Appendix A), which will show how subtle model distinctions and design decisions affect the domain-driven design process.

  The diagram on the top of the next page is a navigation map. It shows the patterns that will be presented in this section and a few of the ways they relate to each other.

  Sharing these standard patterns brings order to the design and makes it easier for team members to understand each other’s work. Using standard patterns also adds to the UBIQUITOUS LANGUAGE, which all team members can use to discuss model and design decisions.

  Developing a good domain model is an art. But the practical design and implementation of a model’s individual elements can be relatively systematic. Isolating the domain design from the mass of other concerns in the software system will greatly clarify the design’s connection to the model. Defining model elements according to certain distinctions sharpens their meanings. Following proven patterns for individual elements helps produce a model that is practical to implement.

  A navigation map of the language of MODEL-DRIVEN DESIGN

  Elaborate models can cut through complexity only if care is taken with the fundamentals, resulting in detailed elements that the team can confidently combine.

  Four. Isolating the Domain

  The part of the software that specifically solves problems from the domain usually constitutes only a small portion of the entire software system, although its importance is disproportionate to its size. To apply our best thinking, we need to be able to look at the elements of our model and see them as a system. We must not be forced to pick them out of a much larger mix of objects, like trying to identify constellations in the night sky. We need to decouple the domain objects from other functions of the system, so we can avoid confusing the domain concepts with other concepts related only to software technology or losing sight of the domain altogether in the mass of the system.

  Sophisticated techniques for this isolation have emerged. This is well-trodden ground, but it is so critical to the successful application of domain-modeling principles that it must be reviewed briefly, from a domain-driven point of view. . . .

  Layered Architecture

  For a shipping application to support the simple user act of selecting a cargo’s destination from a list of cities, there must be program code that (1) draws a widget on the screen, (2) queries the database for all the possible cities, (3) interprets the user’s input and validates it, (4) associates the selected city with the cargo, and (5) commits the change to the database. All of this code is part of the same program, but only a little of it is related to the business of shipping.

  Software programs involve design and code to carry out many different kinds of tasks. They accept user input, carry out business logic, access databases, communicate over networks, display information to users, and so on. So the code involved in each program function can be substantial.

  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 happens because it is the easiest way to make things work, in the short run.

  When the domain-related code is diffused through such a large amount of other code, it becomes extremely difficult to see and to reason 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.

  Creating programs that can handle very complex tasks calls for separation of concerns, allowing concentration on different parts of the design in isolation. At the same time, the intricate interactions within the system must be maintained in spite of the separation.

  There are all sorts of ways a software system might be divided, but through experience and convention, the industry has converged on LAYERED ARCHITECTURES, and specifically a few fairly standard layers. The metaphor of layering is so widely used that it feels intuitive to most developers. Many good discussions of layering are available in the literature, sometimes in the format of a pattern (as in Buschmann et al. 1996, pp. 31–51). The essential principle is that any element of a layer depends only on other elements in the
same layer or on elements of the layers “beneath” it. Communication upward must pass through some indirect mechanism, which I’ll discuss a little later.

  The value of layers is that each specializes in a particular aspect of a computer program. This specialization allows more cohesive designs of each aspect, and it makes these designs much easier to interpret. Of course, it is vital to choose layers that isolate the most important cohesive design aspects. Again, experience and convention have led to some convergence. Although there are many variations, most successful architectures use some version of these four conceptual layers:

  Some projects don’t make a sharp distinction between the user interface and application layers. Others have multiple infrastructure layers. But it is the crucial separation of the domain layer that enables MODEL-DRIVEN DESIGN.

  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.

  Separating the domain layer from the infrastructure and user interface layers allows a much cleaner design of each layer. Isolated layers are much less expensive to maintain, because they tend to evolve at different rates and respond to different needs. The separation also helps with deployment in a distributed system, by allowing different layers to be placed flexibly in different servers or clients, in order to minimize communication overhead and improve performance (Fowler 1996).