Constraints in object-oriented analysis

Abstract

Object-oriented analysis methods can incorporate the concept of constraints to express rules of the problem domain in the specification model, restricting the possible instances of the model. As such, constraints describe properties that must be true at each moment in time for the entire system, without determining how they are to be preserved. The ways in which these constraints are introduced in the model differ from method to method, and even between distinct constraint types in a single method. Different ways in which constraints can be described, are illustrated and compared. Specifying constraints as informal annotations or by operational restrictions is too informal and low level for analysis. According to the properties, importance and influence of the constraint types on the object model, they ought to be described differently. Some constraints, such as connectivity constraints, should best be integrated in existing model concepts to focus on the constraint during the concept definition and as a reminder for these kind of constraints Others, such as attribute value constraints, are best introduced as independent items part of a separate concept grafted on a general model to get a consistent, unambiguous, symmetrical and general applicable constraint description. Yet others, such as relational and existential dependency constraints, should be expressed implicitly by a hierarchical model structure. This approach enriches the object model in such a way that it highlights the logical structure of the problem domain to its right extent.