The Unified Modeling Language User Guide SECOND EDITION [Electronic resources] نسخه متنی

Common Modeling Techniques

Modeling Simple Collaborations

No class stands alone. Rather, each works in collaboration with others to carry out some semantics greater than each individual. Therefore, in addition to capturing the vocabulary of your system, you'll also need to turn your attention to visualizing, specifying, constructing, and documenting the various ways these things in your vocabulary work together. You use class diagrams to represent such collaborations.

To model a collaboration,

• Identify the mechanism you'd like to model. A mechanism represents some function or behavior of the part of the system you are modeling that results from the interaction of a society of classes, interfaces, and other things.

• For each mechanism, identify the classes, interfaces, and other collaborations that participate in this collaboration. Identify the relationships among these things as well.

• Use scenarios to walk through these things. Along the way, you'll discover parts of your model that were missing and parts that were just plain semantically wrong.

• Be sure to populate these elements with their contents. For classes, start with getting a good balance of responsibilities. Then, over time, turn these into concrete attributes and operations.

For example, Figure 8-2 shows a set of classes drawn from the implementation of an autonomous robot. The figure focuses on the classes involved in the mechanism for moving the robot along a path. You'll find one abstract class (Motor) with two concrete children, SteeringMotor and MainMotor. Both of these classes inherit the five operations of their parent, Motor. The two classes are, in turn, shown as parts of another class, Driver. The class PathAgent has a one-to-one association to Driver and a one-to-many association to CollisionSensor. No attributes or operations are shown for PathAgent, although its responsibilities are given.

There are many more classes involved in this system, but this diagram focuses only on those abstractions that are directly involved in moving the robot. You'll see some of these same classes in other diagrams. For example, although not shown here, the class PathAgent collaborates with at least two other classes (Environment and GoalAgent) in a higher-level mechanism for managing the conflicting goals the robot might have at a given moment. Similarly, also not shown here, the classes CollisionSensor and Driver (and its parts) collaborate with another class (FaultAgent) in a mechanism responsible for continuously checking the robot's hardware for errors. By focusing on each of these collaborations in different diagrams, you provide an understandable view of the system from several angles.

Modeling a Logical Database Schema

Many of the systems you'll model will have persistent objects, which means that they can be stored in a database for later retrieval. Most often, you'll use a relational database, an object-oriented database, or a hybrid object/relational database for persistent storage. The UML is well-suited to modeling logical database schemas, as well as physical databases themselves.

The UML's class diagrams are a superset of entity-relationship (E-R) diagrams, a common modeling tool for logical database design. Whereas classical E-R diagrams focus only on data, class diagrams go a step further by permitting the modeling of behavior as well. In the physical database, these logical operations are generally turned into triggers or stored procedures.

To model a schema,

• Identify those classes in your model whose state must transcend the lifetime of their applications.

• Create a class diagram that contains these classes. You can define your own set of stereotypes and tagged values to address database-specific details.

  Stereotypes are discussed in Chapter 6 .

• Expand the structural details of these classes. In general, this means specifying the details of their attributes and focusing on the associations and their multiplicities that relate these classes.

• Watch for common patterns that complicate physical database design, such as cyclic associations and one-to-one associations. Where necessary, create intermediate abstractions to simplify your logical structure.

• Consider also the behavior of these classes by expanding operations that are important for data access and data integrity. In general, to provide a better separation of concerns, business rules concerned with the manipulation of sets of these objects should be encapsulated in a layer above these persistent classes.

• Where possible, use tools to help you transform your logical design into a physical design.

Note

Logical database design is beyond the scope of this book. The focus here is simply to show how you can model schemas using the UML. In practice, you'll end up using stereotypes tuned to the kind of database (relational or object-oriented) you are using.

Figure 8-3 shows a set of classes drawn from an information system for a school. This figure expands upon an earlier class diagram, and you'll see the details of these classes revealed to a level sufficient to construct a physical database. Starting at the bottom-left of this diagram, you will find the classes named Student, Course, and Instructor. There's an association between Student and Course, specifying that students attend courses. Furthermore, every student may attend any number of courses, and every course may have any number of students.

This diagram exposes the attributes of all six of these classes. Notice that the types of all the attributes are primitive types. When you are modeling a schema, you'll generally want to model the relationship to any nonprimitive types using an explicit association rather than an attribute.

Two of these classes (School and Department) expose several operations for manipulating their parts. These operations are included because they are important to maintain data integrity (adding or removing a Department, for example, will have some rippling effects). There are many other operations that you might consider for these and the other classes, such as querying the prerequisites of a course before assigning a student. These function more as business rules than as operations for database integrity, so they are best placed at a higher level of abstraction than this schema.

Forward and Reverse Engineering

Modeling is important, but you have to remember that the primary product of a development team is software, not diagrams. Of course, the reason for creating models is to be able to deliver software that satisfies the evolving goals of its users and the business at the right time. For this reason, it's important that the models you create and the implementations you deploy map to one another and do so in a way that minimizes or even eliminates the cost of keeping your models and your implementation in sync with one another.

For some uses of the UML, the models you create will never map to code. For example, if you are modeling a business process using activity diagrams, many of the activities you model will involve people, not computers. In other cases, you'll want to model systems whose parts are, from your level of abstraction, just a piece of hardware (although at another level of abstraction, it's a good bet that this hardware contains an embedded computer and software).

In most cases though, the models you create will map to code. The UML does not specify a particular mapping to any object-oriented programming language, but the UML was designed with such mappings in mind. This is especially true for class diagrams, whose contents have a clear mapping to all the industrial-strength object-oriented languages, such as Java, C++, Smalltalk, Eiffel, Ada, ObjectPascal, and Forte. The UML was also designed to map to a variety of commercial object-based languages, such as Visual Basic.

Note

The mapping of the UML to specific implementation languages for forward and reverse engineering is beyond the scope of this book. In practice, you'll end up using stereotypes and tagged values tuned to the programming language you are using.

Forward engineering is the process of transforming a model into code through a mapping to an implementation language. Forward engineering results in a loss of information, because models written in the UML are semantically richer than any current object-oriented programming language. In fact, this is a major reason why you need models in addition to code. Structural features, such as collaborations, and behavioral features, such as interactions, can be visualized clearly in the UML, but not so clearly from raw code.

To forward engineer a class diagram,

• Identify the rules for mapping to your implementation language or languages of choice. This is something you'll want to do for your project or your organization as a whole.

• Depending on the semantics of the languages you choose, you may want to constrain your use of certain UML features. For example, the UML permits you to model multiple inheritance, but Smalltalk permits only single inheritance. You can choose to prohibit developers from modeling with multiple inheritance (which makes your models language-dependent), or you can develop idioms that transform these richer features into the implementation language (which makes the mapping more complex).

• Use tagged values to guide implementation choices in your target language. You can do this at the level of individual classes if you need precise control. You can also do so at a higher level, such as with collaborations or packages.

• Use tools to generate code.

Figure 8-4 illustrates a simple class diagram specifying an instantiation of the chain of responsibility pattern. This particular instantiation involves three classes: Client, EventHandler, and GUIEventHandler. The classes Client and EventHandler are abstract, whereas GUIEventHandler is concrete. EventHandler has the usual operation expected of this pattern (handleRequest), although two private attributes have been added for this instantiation.

All of these classes specify a mapping to Java, as noted in their stereotype. Forward engineering the classes in this diagram to Java is straightforward, using a tool. Forward engineering the class EventHandler yields the following code.

public abstract class EventHandler {
EventHandler successor;
private Integer currentEventID;
private String source;
EventHandler() {}
public void handleRequest() {}
}

Reverse engineering is the process of transforming code into a model through a mapping from a specific implementation language. Reverse engineering results in a flood of information, some of which is at a lower level of detail than you'll need to build useful models. At the same time, reverse engineering is incomplete. There is a loss of information when forward engineering models into code, and so you can't completely recreate a model from code unless your tools encode information in the source comments that goes beyond the semantics of the implementation language.

To reverse engineer a class diagram,

• Identify the rules for mapping from your implementation language or languages of choice. This is something you'll want to do for your project or your organization as a whole.

• Using a tool, point to the code you'd like to reverse engineer. Use your tool to generate a new model or modify an existing one that was previously forward engineered. It is unreasonable to expect to reverse engineer a single concise model from a large body of code. You need to select portion of the code and build the model from the bottom.

• Using your tool, create a class diagram by querying the model. For example, you might start with one or more classes, then expand the diagram by following specific relationships or other neighboring classes. Expose or hide details of the contents of this class diagram as necessary to communicate your intent.

• Manually add design information to the model to express the intent of the design that is missing or hidden in the code.