Avoiding Pitfalls in Data Modeling - High-Performance Web Databases: Design, Development, and Deployment - page 276

an additional entity videocassette. A video store may carry 500 titles and
have exactly two copies of each title. Using the video title entity allows the
business to determine whether a given title is in its catalog but not whether
a particular cassette has been returned. Using the video title entity
together with the videocassette entity provides greater capability to
answer business questions.

Mixing concrete and abstract objects is an easy trap to fall into, in part
because in the English language, a given word can be used in more than one
context. Some authors refer to the problem as modeling at different levels
of abstraction. To differentiate the levels, the attributes associated with the
proposed entities must be specified.

Relationship

A relationship is an association, of interest to the enterprise, that can
exist between two entities. The specification of a relationship includes the
names of the entities involved, the semantics of the relationship between
them, and the cardinality and optionality involved. Relationship is an
essential concept in data modeling because it is one of the key vehicles by
which the business rules are represented.

The most common way to identify relation-
ships is through business requirement interviews with users. Users could
be asked directly about the relationship between entities and could be
asked to define entities for the modeler as well-not only in terms of what
the entity is but in terms of how it interacts with other entities. For exam-
ple, a customer might be defined as a company or an individual that buys
goods or services from the enterprise, on the basis of the terms and condi-
tions covered in a master contract. Another way to discover relationships
is to examine existing documents, screens, reports, and forms to uncover
what some of the existing relationships are, on the basis of inferences that
could be made from existing business practices. The pitfalls of missing
relationships can be minimized by asking users what business questions
must be answered in relation to two or more entities.

Relationship Identification.

It is a fallacy to think that, for example,
because a physical purchase order form contains vendor name, vendor
warehouse address, office address, product, and product price that the
entities containing these data elements have to be directly related to the
purchase order entity. It would be helpful instead to recognize that the
physical purchase order form is really a data view, which is made up of
the different entities (e.g., vendor, product, and address), as well as the
purchase order entity itself. The various entities may be related, but all
the entities are not by implication related directly to the purchase order
entity itself.

Misconstrued Relationships.