Constraint specifications generating rules in semantic models ORM and ERMM

Any application of database (DB) technology needs DB scheme designing. If an application domain (AD) is complex, its definition in language of the database management system (DBMS) is not easy task. In such cases, designers usually resort to a semantic design technique providing initial formalization of AD information within the framework of a semantic model and its subsequent transformation in DBMS data model. Widespread semantic models, such as ER and EER, possess insufficiently expressive means of AD semantics description. So, such DB scheme contains only definitions of data structure and declarative integrity constraints, which can be presented in DBMS Data Definition Language. In this case, the designer has to subsequently append the missing elements of DB scheme (views, triggers, etc.) in the language of DB server and then again to analyze AD. Thus, very important task is entrusted to semantic data models; this is an initial formal presentation of AD information. This task becomes complicated if we want to obtain high-level but complete description of AD semantics. Among three data model components (structure data rules, integrity constraint (IC) rules, data operations) an IC instruments make an important contribution to solution of the complete AD scheme design task. Although constraint role in semantics fixation task is rare recollected; all people say about their participation in solution of actual DB state support problem. Tsichritzis and Lochovsky define IC purpose in the following way: «Constraints are introduced in data models with a view to enhancement of their semantic expressiveness and to enhancement of data integrity support. The first aspect concerns adequacy of real world reflection in data scheme and the second - ability of DBMS to guarantee correspondence of database states to requirements, expressed by constraints». Therefore, the modern situation with this component of semantic data models is of interest. In this article, the facilities of integrity constraint declarative definition are analyzed by the examples of two more expressive semantic data models - «Object - Role» (OR-model) and «Entity - Relationship - Mapping» (ERM-model). Authors of these models understand the necessity of construction of an expressive semantic data model and so they attend to constraint definition rules, with help of which AD business rules are declared in data scheme. Heretofore, this aspect was insufficiently developed and manual realizations of many business rules in DBMS programming language were required. If complete high-level description of AD semantics is a success, a question about an automatization of transformation such schemata to complete DBMS schemata may bring up. Following OR-constraint classes are the basis of the article's content: - internal and external role uniqueness constraints, - mandatory role constraints, - inclusive-or constraints, - exclusion constraints, - exclusive-or constraints, - set comparison constraints (subset, equality and exclusion), - subtyping constraints, - occurrence frequencies constraints, - ring constraints. Each IC class is illustrated by example from Halpin's book; the same example is presented in ERM notation. Thus, underlying model principles are clearly demonstrated. Analysis, carried out in article, shows that described model's authors chose different ways to solution of the semantic expressiveness enhancement problem. Feature of Halpin's model IC instruments is that special term and graphic primitive are contrived for any useful property of data. Model unhandiness and complexity of its use are results of such approach. Unlike OR-model ERM-model offers limited set of universal constructions, with help of which many (if not all) data relation rules can be presented. At least all kinds of OR-model IC can be expressed by ERM IC tools. The ERM basis is formed by universal concepts of logic and mathematics.

Keywords

Authors

References