Split (1)

test · 3k rows

text stringlengths 16 313
These identiﬁers can beused as arguments to subsequent messages.
In other cases, it may be the name of an event that israised on the receiving object.
In normal practice withprocedural overloading, both the message name and theargument list types are required to identify an operation.
The parentheses can be used even if the list isempty.
The following are samples of control message label syntax.
Instead of text expressions for arguments and return values, data tokens may beshown near a message (Figure 13-123).
A token is a small circle labeled with theargument expression or return value name.
It has a small arrow on it that pointsalong the message (for an argument) or opposite the message (for a return value).
Tokens represent arguments and return values.
The choice of text syntax or tokensis a presentation option, but text is more compact and is recommended for mostpurposes.
The syntax of messages may be expressed in the syntax of a programming lan-guage, such as C++ or Smalltalk.
All the expressions on a single diagram shoulduse the same syntax, however.
A class whose instances are classes.
Metaclasses are typically used to constructmetamodels.
A model that deﬁnes the language for expressing a metamodel.
The relationshipbetween a meta-metamodel and a metamodel is analogous to the relationship be-tween a metamodel and a model.
This level of indirection is usually relevant onlyto tool builders, database builders, and the like.
UML is deﬁned in terms of a meta-metamodel, called the Meta-Object Facility (MOF).
A model that deﬁnes the language for expressing a model; an instance of a meta-metamodel.
The UML metamodel deﬁnes the structure of UML models.
A generic term for all entities in a metamodeling language.
A term grouping relationships that connect descriptors to their instances.
Theseinclude the instance relationship and the powertype relationship.
A method is an implementation of an operation.
The language must be matched to the purpose,of course.
A human language, for instance, may be adequate for early analysis butnot suitable for code generation.
An operation declaration implies the presence of a method unless the operationis declared as abstract.
In a generalization hierarchy, each repeated declaration ofthe operation implies a new method that overrides any inherited method of thesame operation.
Two declarations represent the same operation if their signaturesmatch.
Note that a method is an executable procedure—an algorithm—not simply aspeciﬁcation of results.
A before-and-after speciﬁcation is not a method, for exam-ple.
A method is a commitment to implementation and addresses issues of algo-rithm, computational complexity, and encapsulation.
In some respects, a method may have stricter properties than its operation.
But ifthe operation is a query, then the method must be a query.
Similarly, a methodmay strengthen the concurrency property.
A sequential operation may be imple-mented as a guarded or concurrent method.
In these cases, the method is consis-tent with the declarations of its operation.
If the operation is inherited, the method can beshown by repeating the operation declaration in normal (nonitalic) text to show aconcrete operation.
Method on nonabstract operationShapedraw ( )
A semantically complete abstraction of a system.
A model is a more or less complete abstraction of a system from a particular view-point.
It is complete in the sense that it fully describes the system or entity, at thechosen level of precision and viewpoint.
Different models provide mostly inde-pendent viewpoints that can be manipulated separately.
A model may comprise a containment hierarchy of packages in which the top-level package corresponds to the entire system.
The contents of a model are thetransitive closure of its containment (ownership) relationships from top-levelpackages to model elements.
A model may also include relevant parts of the system’s environment, repre-sented, for example, by actors and their interfaces.
In particular, the relationshipof the environment to the system elements may be modeled.
A system and its envi-ronment form a larger system at a higher level of scope.
Therefore, it is possible torelate elements at various levels of detail in a smooth way.
Elements in different models do not directly affect each other, but they oftenrepresent the same concepts at different levels of detail or stages of development.
Therefore, relationships among them, such as trace and reﬁnement, are importantto the development process itself and often capture important design decisions.
A model can be shown as a package with the stereotype «model».
There is little no-tational detail to show about models, however.
Tools can show lists of models, butmodels have few relationships among themselves.
Most useful is the ability totraverse from a model name to its top package or to a map of its overall contents.
No one view of a system, or indeed no system itself, is ever complete in and of it-self.
There are always connections to the wider world, and a model always fallsshort of reality.
Therefore, the concept of a closed model is always an approxima-tion in which arbitrary lines must be drawn for practical work.
A UML model is represented as a package hierarchy that emphasizes one view ofa system.
An element that is an abstraction drawn from the system being modeled.
Contrastwith presentation element, which is a (generally visual) presentation of one ormore modeling elements for human interaction.
All elements that have semantics are model elements, including real-world con-cepts and computer-system implementation concepts.
Graphic elements whosepurpose is to visualize a model are presentation elements.
They are not model ele-ments, as they do not add semantics to the model.
Model elements may have names, but the use and constraints on names vary bykind of model element and are discussed with each kind.
Each model element be-longs to a namespace appropriate to the kind of element.
All model elements mayhave the following attached properties.
The tag is aname that identiﬁes the meaning of the value.
Constraints are restrictions that are expressed as
The stereotype does not alter the struc-ture of the base class, but it may add constraints andtagged values that apply to the model elements bearingthe stereotype.
In addition, model elements may participate in dependency relationships.
See Chapter 14, Standard Elements, for a list of predeﬁned tags, constraints, andstereotypes.
That aspect of a model dealing with the organization of the model itself into struc-tured parts—namely, packages, subsystems, and models.
Refers to something that occurs during a modeling activity of the software devel-opment process.
Usage note: When discussingobject systems, it is often important to distinguish between modeling-time andrun-time concerns.
See also development process, stages of modeling.
A software unit of storage and manipulation.
Modules include source code mod-ules, binary code modules, and executable code modules.
The word does notcorrespond to a single UML construct, but rather includes several constructs.
A classiﬁer role that denotes a set of objects rather than a single object.
See also classiﬁer role, collaboration, message.
A multiobject is a classiﬁer role that denotes a set of objects, usually the set of ob-jects on the many side of an association.
A multiobject is used within a collabora-tion to show operations that address the entire set of objects as a unit rather than asingle object in it.
For example, an operation to ﬁnd an object within a set operateson the entire set, not on an individual object.
The underlying static model is unaf-fected by this grouping.
A multiobject is shown as two rectangles in which the top rectangle is shiftedslightly vertically and horizontally to suggest a stack of rectangles (Figure 13-125).
A message arrow to the multiobject symbol indicates a message to the set of ob-jects—for example, a selection operation to ﬁnd an individual object.
To perform an operation on each object in a set of associated objects requirestwo messages: an iteration to the multiobject to extract links to the individual ob-jects, then a message sent to each object using the (temporary) link.
This may beelided on a diagram by combining the messages into one that includes an iterationand an application to each object.
The target rolename takes a many indicator (*)to show that many links are implied.
An object from the set is shown as a normal object symbol, but it may be at-tached to the multiobject symbol using a composition link to indicate that it ispart of the set.
A message arrow to the simple object symbol indicates a message toan individual object.
Typically, a selection message to a multiobject returns a reference to an individ-ual object, to which the original sender then sends a message.
A semantic variation of generalization in which an object may belong directly tomore than one class.
This is a semantic variation point under which an object may be a direct instanceof more than one class.
When used with dynamic classiﬁcation, objects may ac-quire and lose classes during run time.
This allows classes to be used to representtemporary roles an object may play.
Although multiple classiﬁcation matches logic and everyday discourse well, itcomplicates implementation of a programming language and is not supported bythe popular programming languages.
A semantic variation point of generalization in which an element may have morethan one parent.
This is the default assumption within UML and is necessary forproper modeling of many situations, although modelers may choose to restrict itsuse for certain kinds of elements.

Subsets and Splits

No community queries yet

The top public SQL queries from the community will appear here once available.