Task: Use-Case Analysis
This task describes how to develop an Analysis-level Use-Case Realization from a Use-Case.
Purpose
  • To identify the classes which perform a use case's flow of events
  • To distribute the use case behavior to those classes, using analysis use-case realizations
  • To identify the responsibilities, attributes and associations of the classes
  • To note the usage of architectural mechanisms
Relationships
Steps
Create Analysis Use-Case Realization
Purpose To create the modeling element used to express the behavior of the use case. 

Use Cases form the central focus of most of the early analysis and design work. To enable the transition between Requirements-centric tasks and Analysis/Design-centric tasks, the Work Product: Use-Case Realization serves as a bridge, providing a way to trace behavior in the Analysis and Design Models back to the Use-Case Model, as well as organizing collaborations around the Use Case concept.

If one does not already exist, create a Analysis Use-Case Realization in the Analysis Model for the Use Case.  The name for the Analysis Use-Case Realization should be the same as the associated Use Case, and a "realizes" relationship should be established from the analysis use-case realization to its associated use case.

For more information on use-case realizations, see Guideline: Use-Case Realization.

Supplement the Use-Case Description
Purpose To capture additional information needed in order to understand the required internal behavior of the system that might be missing from the use-case description written for the customer of the system. 

The description of each use case is not always sufficient for finding analysis classes and their objects. The customer generally finds information about what happens inside the system uninteresting, so the use-case descriptions may leave such information out. In these cases, the use-case description reads like a 'black-box' description, in which internal details on what the system does in response to an actor's actions is either missing or very summarily described. To find the objects which perform the use case, you need to have the 'white box' description of what the system does from an internal perspective.

Example

In the case of an Automated Teller Machine (ATM), the customer of the system might prefer to say

"The ATM validates the Bank Customer's card."

To describe the user authentication behavior of the system. While this might be sufficient for the customer, it gives us no real idea of what actually happens inside the ATM to validate the card.

In order to form an internal picture of how the system works, at a sufficient level of detail to identify objects, we might need additional information. Taking the ATM card validation activity as an example, the expanded description would read as:

"The ATM sends the customer's account number and the PIN to the ATM Network to be validated. The ATM Network returns success if the customer number and the PIN match and the customer is authorized to perform transactions, otherwise the ATM Network returns failure."

This level of detail provides a clear idea of what information is required (account number and PIN) and who is responsible for the authentication (the ATM Network, an actor in the Use Case model). From this information, we can identify two potential objects (a Customer object, with attributes of account number and PIN, and an ATM Network Interface) as well as their responsibilities.

Examine the use-case description to see if the internal behavior of the system is clearly defined. The internal behavior of the system should be unambiguous, so that it is clear what the system must do. It is not necessary to define the elements within the system (objects) that are responsible for performing that behavior - just a clear definition of what needs to be done.

Sources of information for this detail include domain experts who can help define what the system needs to do. A good question to ask, when considering a particular behavior of the system, is "what does it mean for the system to do that thing?". If what the system does to perform the behavior is not well defined enough to answer that question, there is likely more information that needs to be uncovered.

The following alternatives exist for supplementing the description of the Flow of Events:

  • Do not describe it at all. This might be the case if you think the interaction diagrams are self-explanatory, or if the Flow of Events of the corresponding use case provides a sufficient description.
  • Supplement the existing Flow of Event description. Add supplementary descriptions to the Flow of Events in areas where the existing text is unclear about the actions the system should take.
  • Describe it as a complete textual flow, separate from the "external" Use Case Flow of Events description. This is appropriate in cases where the internal behavior of the system bears little resemblance to the external behavior of the system. In this case, a completely separate description, associated with the analysis use-case realization rather than the use case, is warranted.
Find Analysis Classes from Use-Case Behavior
Purpose To identify a candidate set of model elements (analysis classes) which will be capable of performing the behavior described in use cases. 

Finding a candidate set of analysis classes is the first step in the transformation of the system from a mere statement of required behavior to a description of how the system will work. In this effort, analysis classes are used to represent the roles of model elements which provide the necessary behavior to fulfill the functional requirements specified by use cases and the non-functional requirements specified by the supplemental requirements. As the project focus shifts to design, these roles evolve a set of design elements which realize the use cases.

The roles identified in Use-Case Analysis primarily express behavior of the upper-most layers of the system-application-specific behavior and domain specific behavior. Boundary classes and control classes typically evolve into application-layer design elements, while entity classes evolve into domain-specific design elements. Lower layer design element typically evolve from the analysis mechanisms which are used by the analysis classes identified here.

The technique described here uses three different perspectives of the system to drive the identification of candidate classes. The three perspectives are that of the boundary between the system and its actors, the information the system uses, and the control logic of the system. The corresponding class stereotypes, boundary, entity and control, are conveniences used during Analysis that disappear in Design.

Identification of classes means just that: they should be identified, named, and described briefly in a few sentences.

For more information on identification of analysis classes, see Guideline: Analysis Class. For more information on analysis use-case realizations, see Guideline: Use-Case Realization.

If particular analysis mechanisms and/or analysis patterns have been documented in the project-specific guidelines, these should be used as another source of "inspiration" for the analysis classes.

Distribute Behavior to Analysis Classes
Purpose To express the use-case behavior in terms of collaborating analysis classes. To determine the responsibilities of analysis classes. 

For each independent sub-flow (scenario):

  • Create one or more interaction (communication or sequence) diagrams. At least one diagram is usually needed for the main flow of events of the use case, plus at least one diagram for each alternate/exceptional flow. Separate diagrams are usually needed for sub-flows which have complex timing or decision points, or to simplify complex flows which are too long to grasp easily in one diagram.
  • Identify the analysis classes responsible for the required behavior by stepping through the flow of events of the scenario, ensuring that all behavior required by the use case is provided by the analysis use-case realization.
  • Illustrate interactions between analysis classes in the interaction diagram. The interaction diagram should also show interactions of the system with its actors (the interactions should begin with an actor, since an actor always invokes the use case).
  • Include classes that represent the control classes of used use-cases. (Use a separate interaction diagram for each extending use-case, showing only the variant behavior of the extending use case.)

an example communication diagram

A communication diagram for the use case Receive Deposit Item.

If particular analysis mechanisms and/or analysis patterns have been documented in the project-specific guidelines, these should be reflected in the allocation of responsibility and resulting interaction diagrams.

Describe Responsibilities
Purpose To describe the responsibilities of a class of objects identified from use-case behavior. 

A responsibility is a statement of something an object can be asked to provide. Responsibilities evolve into one (but usually more) operations on classes in design; they can be characterized as:

  • the actions that the object can perform
  • the knowledge that the object maintains and provides to other objects

Each analysis class should have several responsibilities; a class with only one responsibility is probably too simple, while one with a dozen or more is pushing the limit of reasonability and should potentially be split into several classes.

That all objects can be created and deleted goes without saying; don't restate the obvious unless the object performs some special behavior when it is created or deleted. (Some objects cannot be removed if certain relationships exist.)

Finding Responsibilities

Responsibilities are derived from messages in interaction diagrams. For each message, examine the class of the object to which the message is sent. If the responsibility does not yet exist, create a new responsibility that provides the requested behavior.

Other responsibilities will derive from non-functional requirements. When you create a new responsibility, check the non-functional requirements to see if there are related requirements which apply. Either augment the description of the responsibility, or create a new responsibility to reflect this.

Documenting Responsibilities

Responsibilities are documented with a short (up to several words) name for the responsibility, and a short (up to several sentences) description. The description states what the object does to fulfill the responsibility, and what result is returned when the responsibility is invoked.

Describe Attributes and Associations
Purpose To define the other classes on which the analysis class depends.
To define the events in other analysis classes that the class must know about.
To define the information that the analysis class is responsible for maintaining. 

In order to carry-out their responsibilities, classes frequently depend on other classes to supply needed behavior. Associations document the inter-class relationships and help us to understand class coupling; better understanding of class coupling, and reduction of coupling where possible, can help us build better, more resilient systems.

The following steps define the attributes of classes and the associations between classes:

Define Attributes

Attributes are used to store information by a class. Specifically, attributes are used where the information is:

  • Referred to "by value"; that is, it is only the value of the information, not it's location or object identifier which is important.
  • Uniquely "owned" by the object to which it belongs; no other objects refer to the information.
  • Accessed by operations which only get, set or perform simple transformations on the information; the information has no "real" behavior other than providing its value.

If, on the other hand, the information has complex behavior, is shared by two or more objects, or is passed "by reference" between two or more objects, the information should be modeled as a separate class.

The attribute name should be a noun that clearly states what information the attribute holds.

The description of the attribute should describe what information is to be stored in the attribute; this can be optional when the information stored is obvious from the attribute name.

The attribute type is the simple data type of the attribute. Examples include string, integer, number.

Establish Associations between Analysis Classes

Start by studying the links in the interaction diagrams produced in Distribute Behavior to Analysis Classes. Links between classes indicate that objects of the two classes need to communicate with one another to perform the Use Case. Once we start designing the system, these links might be realized in several ways:

  • The object might have "global" scope, in which case any object in the system can send messages to it
  • One object might be passed the second object as a parameter, after which it can send messages to the passed object.
  • The object might have a permanent association to the object to which messages are sent.
  • The object might be created and destroyed within the scope of the operation (i.e. a 'temporary' object)-these objects are considered to be 'local' to the operation.

At this early point in the "life" of the class, however, it is too early to start making these decisions: we do not yet have enough information to make well-educated decisions. As a result, in analysis we create associations and aggregations to represent (and "carry") any messages that must be sent between objects of two classes. (Aggregation, a special form of association, indicates that the objects participate in a "whole/part" relationship (see Guideline: Association and Guideline: Aggregation)).

We will refine these associations and aggregations in the Task: Class Design.

For each class, draw a class diagram which shows the associations each class has to other classes:

an example class diagram, showing associations and aggregations

Example analysis class diagram for part of an Order Entry System

Focus only on associations needed to realize the use cases; don't add association you think "might" exist unless they are required based on the interaction diagrams.

Give the associations role names and multiplicities.

  • A role name should be a noun expressing what role the associated object plays in relation to the associating object.
  • Assume a multiplicity of 0..* (zero to many) unless there is some clear evidence of something else. A multiplicity of zero implies that the association is optional; make sure you mean this; if an object might not be there, operations which use the association will have to adjust accordingly.
  • Narrower limits for multiplicity may be specified (such as 3..8).
  • Within multiplicity ranges, probabilities may be specified. Thus, if the multiplicity is 0..*, is expected to be between 10 and 20 in 85% of the cases, make note of it; this information will be of great importance during design. For example, if persistent storage is to be implemented using a relational database, narrower limits will help better organize the database tables.

Write a brief description of the association to indicate how the association is used, or what relationships the association represents.

Describe Event Dependencies between Analysis Classes

Objects sometimes need to know when an event occurs in some "target" object, without the "target" having to know all the objects which require notification when the event occurs. As a short-hand to show this event-notification dependency, a subscribe-association allows us to express this dependency in a compact, concise way.

A subscribe-association between two objects indicates that the subscribing object will be informed when a particular event has occurred in the subscribed object. A subscribe-association has a condition defining the event that causes the subscriber to be notified. For more information, see Guideline: Subscribe-Association

The conditions of the subscribes-association should be expressed in terms of abstract characteristics, rather than in terms of its specific attributes or operations. In this way, the associating object is kept independent of the contents of the associated entity object, which may well change.

A subscribe-association is needed:

  • if an object is influenced by something that occurs in another object
  • if a new object must be created to deal with some event, for example, when an error occurs, a new window must be created to notify the user
  • if an object needs to know when another object is instantiated, changed or destroyed

The objects which are 'subscribed-to' are typically entity objects. Entity objects are typically passive stores of information, with any behavior generally related to their information-storage responsibilities. Many other objects often need to know when the entity objects change. The subscribe-association prevents the entity object from having to know about all these other objects-they simply 'register' interest in the entity object and are notified when the entity object changes.

Now this is all just 'analysis sleight-of-hand': in design we have to define how exactly this notification works. We might purchase a notification framework, or we might have to design and build one ourselves. But for the moment, simply noting that the notification exists is sufficient.

The direction of the association shows that only the subscribing object is aware of the relation between the two objects. The description of the subscription is entirely within the subscribing object. The associated entity object, in turn, is defined in the usual way without considering that other objects might be interested in its activity. This also implies that a subscribing object can be added to, or removed from, the model without changing the object to which it subscribes.

Reconcile the Analysis Use-Case Realizations
Purpose To reconcile the individual analysis use-case realizations and identify a set of analysis classes with consistent relationships. 

The analysis use-case realizations were developed as a result of analyzing a particular use case.  Now the individual analysis use-case realizations need to be reconciled.  Examine the Analysis Classes and the supporting associations defined for each of the analysis use-case realizations.  Identify and resolve inconsistencies and remove any duplicates.  For example, two different analysis use-case realizations might include an analysis class that is conceptually the same, but since the analysis classes were identified by different Designers, a different name was used. 
Note: Duplication across analysis use-case realizations can be significantly reduced if the Software Architect does a good job defining an initial architecture (see Task: Architectural Analysis).

When reconciling the model elements, it is important to take into consideration their relationships.  If two classes are merged, or one class replaces another, be sure to propagate the original class's relationships to the new class.

The Software Architect should participate in the reconciliation of the analysis use-case realizations, as it requires an understanding of the business context, as well as some foresight of the software architecture and design so that the analysis classes that best represent the problem and solution domains can be selected.

For more information on classes, see Guideline: Analysis Class.

Qualify Analysis Mechanisms
Purpose To identify analysis mechanisms (if any) used by the analysis classes. To provide additional information about how the analysis classes apply the analysis mechanism. 

In this step, the analysis mechanisms that apply to each of the identified analysis classes is examined.

If an analysis class uses one or more analysis mechanisms, additional information captured now will assist the software architect and designers to determine the capabilities required of the architectural design mechanisms. The number of instances of the analysis class, their size, their frequency of access, and their expected life-span are among the important properties that can assist the designers in selecting appropriate mechanisms.

For each analysis mechanism used by an analysis class, qualify the relevant characteristics which need to be considered when selecting appropriate design and implementation mechanisms. These will vary depending on the type of mechanism; give ranges where appropriate, or when there is still much uncertainty. Different architectural mechanisms will have different characteristics, so this information is purely descriptive and need only be as structured as necessary to capture and convey the information. During analysis, this information is generally quite speculative, but capturing has value since conjectural estimates can be revised as more information is uncovered.

The analysis mechanisms used by a class and their associated characteristic need not be captured in a formal way; a note attached to a diagram, or an extension to the description of the class is sufficient to convey the information. The characteristic information at this point in the evolution of the class is quite fluid and speculative, so the emphasis is on capturing expected values rather than on formalizing the definition of the mechanisms.

Example

The characteristics of the persistence mechanism used by a Flight class could be qualified as:

Granularity: 2 to 24 Kbytes per flight

Volume: Up to 100,000

Access frequency:

  • Creation/deletion: 100 per hour
  • Update: 3,000 updates per hour
  • Read: 9,000 access per hour

Example

The characteristics of the persistence mechanism used by a Mission class could be qualified as:

Granularity: 2 to 3 Mbytes per mission

Volume: 4

Access frequency:

  • Creation/deletion: 1 per day
  • Update: 10 per day
  • Read: 100 per hour
Establish Traceability
Purpose To maintain the traceability relationships between the Analysis Model and other models. 

The project's project-specific guidelines specifies what traceability is required for Analysis Model elements.

For example, if there is a separate model of the user interface, then it might be useful to trace screens or other user interface elements in that model to boundary classes in the Analysis Model.

Review the Results
Purpose To verify that the analysis objects meet the functional requirements made on the system.
To verify that the analysis objects and interactions are consistent. 

Conduct a review informally at the end of the workshop, as a synchronization point, as well as the conclusion to the Task: Use-Case Analysis.

Use the checklist(s) for work products output by this task.



Properties
Multiple Occurrences
Event-Driven
Ongoing
Optional
Planned
Repeatable
More Information