An Agile Introduction – The Agile Modeling (AM) Method: (2023)

Figure 1. UML 2.x component diagram.

Figure 2. UML 1.x component diagram.

Diagrams such asFigure 1are often referred to as “wiring diagrams” because they show how the various software components are “wired together” to build your overall application. The lines between components are often referred to as connectors, the implication being that some sort of messaging will occur across the connectors.

I usually draw component diagrams on whiteboards although for both of the examples I’ve used a drawing tool to depict the notation accurately. You can use component diagrams for both logical and physical modeling although I prefer to use them for physical modeling of the software architecture of a system.Figure 1shows the large-scale domain components for the system we’re building, including two user interface components which map to two different applications which we’re building as part of the overall system. This diagram includes both business and technical architecture aspects – the components with theinfrastructureanddatabasestereotypes are clearly technical in nature – and that’s perfectly fine. The important thing is that we’re considering both business and technical aspects in our architecture, not just technical issues, and for whatever reason we’ve chosen to create a single diagram which includes both views.

Interfaces and Ports

Components may both provide and require interfaces. An interface is the definition of a collection of one or more methods, and zero or more attributes, ideally one that defines a cohesive set of behaviors. A provided interface is modeled using the lollipop notation and a required interface is modeled using the socket notation. A port is a feature of a classifier that specifies a distinct interaction point between the classifier and its environment. Ports are depicted as small squares on the sides of classifiers.

For example,Figure 3shows a detailed component diagram which contains three components. There are several interesting features to note about this diagram:

1. Ports can be named, such as theSecurityandDataports on theStudentcomponent.
2. Ports can support unidirectional communication or bi-directional communication. TheStudentcomponent implements three ports, two unidirectional ports and one bi-directional ports. The left-most port is an input port, theSecurityport is an output port, and theDataport is a bi-directional port.
3. TheStudentAdministrationandStudentScheduleinterfaces are application specific and may include overlapping method signatures. I’ve found this approach to be more understandable to the clients of the component as each application team is provided their own specific interface which isn’t encumbered with methods they don’t need.
4. The diagram isn’t “wired” together yet – I haven’t connected theStudentcomponent to the two security components yet.
5. You don’t need to use all of the provided interfaces of a component. My team has decided to use theX-Securitycomponent foraccess controland theMegaEncryptercomponent for encryption. Both of these fictional components are commercial off the shelf packages (COTS) which we’ve purchased. AlthoughX-Securityimplements both security interfaces required byStudentthe other component implements theEncryptioninterface much more efficiently so we’ve decided to use both security components.

Figure 3. Modeling interfaces and ports.

Implementinga Component

So how do you actually build a component? Although there are various strategies to do so, there are several basic principles that you can follow.Figure 4depicts a design for the Student component, depicting it as a UML frame. It’s also common to use “composite structure“, e.g. a rectangle with the component stereotype in the top-right corner, instead of a frame because a component really is a structure composed of smaller elements. Whether you call it a frame, a composite structure, or something else, the diagram is pretty much the same. Interesting points about this diagram are:

1. I simplified the ports to either provide or require a single interface. This enables me to easily and explicitly model the relationships between the ports and the internals of the component.
2. I’ve modeled relationships between ports and internal classes in three different ways: as a stereotyped delegates relationship, as a delegates relationship, and as a realizes relationship. A delegates relationship is a line with an open arrowhead on it and a realizes relationship is a dashed arrow with a closed arrowhead. Because the delegates notation is exactly the same notation that is used for unidirectional associations there is an opportunity for confusion – as a result I recommend indicating the stereotype on the relationship to make it clear what you mean. The realizes notation, for exampleAdministrationFacaderealizes the port containing theStudentAdministrationinterface, used to be my preferred approach because in my mind ports are logical modeling constructs that are realized by physical constructs such as classes. However, the delegates association has the advantage that it indicates the flow of communication and as a result seems to be easier to understand.
3. TheDataandSecurityclasses use the same names as the corresponding ports.
4. Classes such asAdministrationFacade,ScheduleFacade,StudentData,Data, andSecurityimplement the Façade design pattern (Gamma et. al. 1995). The basic idea is that they implement the public operations required by the interfaces, operations that typically just delegate messages to the appropriate classes. Together theAdministrationFacade,ScheduleFacade, andStudentDataclasses implement the public interface of theStudentcomponent.StudentData,Data, andSecuritywrap access to external components so that the internal classes are not directly coupled to other physical components.
5. Expect to adapt incoming/outgoing interactions back and forth between data-oriented and object-oriented messages. An incoming message may be implemented as a web service which takes XML as a parameter and returns XML as the result. The internal objects within the component, on the other hand, need messages sent to them with either objects or data as parameters and return values. The implication is that you need to marshal the data and objects back and forth between each other, something addressed by theAdapterdesign pattern (Gamma et al., 1995).
6. Another way to implement the public interface would have been to implement a single façade class calledStudentComponentwhich implements the required public interfaces and delegates appropriately.
7. When designing theStudentDataclass the team realized that it needed to work with our existingXMLProcessorcomponent, therefore we added the connection to this component.

Figure 4. Designing a component.

Figure 3makes it obvious that building components is costly. Creating theStudentcomponent as shown inFigure 3doesn’t make much sense – I’ve added five new classes to support two domain classes, a clear case of overbuilding. This approach would make sense if there was twenty classes, and would make a lot of sense for fifty domain classes, because the additional five classes reduce the coupling within your system while at the same time implement a large-scale, reusable domain component. The point is that you should only take a component-based approach when the benefit of doing so outweighs the additional cost.

CreatingComponent Diagrams

There are two fundamental strategies for developing a component model, either top down or bottom up. Given the choice I prefer the top-down approach because it provides a good mechanism for identifying the “software landscape” early in the project, something that is particularly important for teams comprised of several subteams because you want to work towards the same vision. Unfortunately a top-down approach suffers from the tendency to promote over-architecting, and hence over-building, of your system. For exampleFigure 1calls outSecurityandPersistencecomponents but you might not yet need anything even remotely that complicated. It would be a serious mistake to focus on building these two components instead of implementing actual business functionality that your stakeholders actually need.

A second way to develop component models is from the bottom up. I’ll do this when we have an existing collection of classes that have been developed and we decide to componentize our design. Componentizing is often done to rescue reusable functionality out of an existing application or to split an application up so it can be easily dispersed between subteams. When I’m componentizing an existing object design I’ll often iterate through the following steps:

1. Keep components cohesive. A component should implement a single, related set of functionality. This may be the user interface logic for a single user application, business classes comprising a large-scale domain concept, or technical classes representing a common infrastructure concept.
2. Assign user interface classes to application components. User interface classes, those that implement screens, pages, or reports, as well as those that implement “glue logic” such as identifying which screen/page/”� to display should be placed in components with theapplicationstereotype. In Java these types of classes would include Java Server Pages (JSPs), servlets, and screen classes implemented via user interface class libraries such as Swing.
3. Assign technical classes to infrastructure components. Technical classes, such as those that implement system-level services such as security, persistence, or middleware should be assigned to components which have theinfrastructurestereotype.
4. Define class contracts. A class contract is any method that directly responds to a message sent from other objects. For example, the contracts of theSeminarclass likely include operations such asenrollStudent()anddropStudent(). For the purpose of identifying components, you can ignore all the operations that aren’t class contracts because they don’t contribute to communication between objects distributed in different components.
5. Assign hierarchies to the same component. 99.9% of the time I find that it makes sense to assign all of the classes of a hierarchy, either an inheritance hierarchy or a composition hierarchy, to the same component.
6. Identify domain components. A domain component is a set of classes that collaborate among themselves to support a cohesive set of contracts. The basic idea is that classes, and even other domain components, are able to send messages to domain components either to request information or to request an action be performed. On the outside, domain components appear simple, actually they appear like any other type of object but, on the inside, they are often quite complex because they encapsulate the behavior of several classes. A key goal is you want to organize your design into several components in such a way as to reduce the amount of information flowing between them. Any information passed between components, either in the form of messages or the objects that are returned as the result of a message send, represents potential traffic on your network (if the components are deployed to different nodes). Because you want to minimize network traffic to reduce the response time of your application, you want to design your domain components in such a way that most of the information flow occurs within the components and not between them.
7. Identify the “collaboration type” of business classes. To determine which domain component a business class belongs to you need to analyze the collaborations it is involved with to determine its distribution type. A server class is one that receives messages, but doesn’t send them. A client class is one that sends messages, but doesn’t receive them. A client/server class is one that both sends and receives messages. Once you have identified the distribution type of each class, you are in a position to start identifying potential domain components.
8. Server classes belong in their own component.Pure server classes belong in a domain component and often form their own domain components because they are the “last stop” for message flow within an application.
9. Merge a component into its only client.If you have a domain component that is a server to only one other domain component, you may decide to combine the two components.
10. Pure client classes don’t belong in domain components.Client classes don’t belong in a domain component because theyonly generate messages, they don’t receive them, whereas the purpose of a domain component is to respond to messages. Therefore, client classes have nothing to add to the functionality offered by a domain component and very likely belong in an application component instead.
11. Highly coupled classes belong in the same component.When two classes collaborate frequently, this is an indication they should be in the same domain component to reduce the network traffic between the two classes. This is especially true when that interaction involves large objects, either passed as parameters or received as return values. By including them in the same domain component you reduce the potential network traffic between them. The basic idea is that highly coupled classes belong together.
12. Minimize the size of the message flow between components.Client/serverclasses belong in a domain component, but there may be a choice as to which domain component they belong to. This is where you need to consider issues such as the information flow going into and out of the class. Communication within a component will often be simple message sends between objects in memory, communication between components may require an expensive marshalling effort in which a message and its parameters are converted to data, transmitted, and then converted back into a message again.
13. Define component contracts. Each component will offer services to its clients, each such service is a component contract.

Table 1summarizes several design principles presented in Agile Software Development (Martin, Newkirk, Koss 2003) for improving the quality of packages or components. I present them here because I find them of greatest value when it comes to component modeling.

Table 1. Component design principles.

Remaining Agile

My most successful use of component models was with a team where we drew a diagram similar to, albeit a much larger one with over twenty components, on a whiteboard. This whiteboard was situated in the team work area where everyone could see the board. We developed the diagram early in theinitiativeand updated it as required throughout the project. We kept it on the board because it provided a high-level map of the architecture of our software, a map that we used from time to time as we worked and more importantly engendered many interesting conversations regarding the overall system design.

There are several advantages to components that promote agility. First, components are reusable building blocks from which you can build software, increasing your productivity as a developer. Second, components can improve your testing productivity because they can be treated as elements which you can black-box unit and integration test. Testing is discussed in detail inAgile Testing and Quality Strategies.

Source

This artifact description is excerpted from Chapter 10 ofThe Object Primer 3rd Edition: Agile Model Driven Development with UML 2.

Translations

• Japanese

Disclaimer

The notation used in these diagrams, particularly the hand drawn ones, may not conform perfectly to the current version of the UML for one or more of reasons:

• The notation may have evolved from when I originally developed the diagrams. The UML evolves over time, and I may not have kept the diagrams up to date.
• I may have gotten it wrong in the first place. Although these diagrams were thoroughly reviewed for the book, and have been reviewed by thousands of people online since then, an error may have gotten past of us. We’re only human.
• I may have chosen to apply the notation in “non-standard” ways. An agile modeler is more interested in created models which communicate effectively than in conforming to notation rules set by a committee.
• It likely doesn’t matter anyway, because themodeling tool(s)that you’re using likely won’t fully support the current version of the UML notation perfectly anyway. Bottom line is that you’re going to be constrained by your tools anyway.

If you’re really concerned about the nuances of “official” UML notation then read the current version of theUML specification.