Problem of data scheme consistency and its solution in the ERM repository

The task of implementing a mechanism for verifying consistency of schemes is one of the main tasks in the development of a data scheme repository. It is important that at any time the data stored in the repository are complete and consistent. To prevent erroneous artifacts created in CASE system, it provides for appropriate checks of the analytical and design documents formed in the repository. In attempts to increase the degree of automation of design and the widening of the spectrum of generated, executable artifacts, the creators of the ERM model significantly expanded the descriptive possibilities allowing, with a limited set of concepts, to determine in the scheme all the structural elements and practically all the business rules of the application domain (AD). In order to simplify the work of designers, along with a basic set of concepts (object and mapping) that provide the entire semantic power of the model, derivative concepts are introduced that allow us to operate with the traditional terms - entity, relationship, attribute. When simple concepts are enough to describe the semantics of AD, they are used, when their descriptive abilities are not enough, the designer uses the basic concepts for a more accurate transfer of semantics. In this case, it becomes possible to have simultaneously in the same scheme different forms of representing the phenomena of AD, in the language of basic and derived terms. The desire to provide storage in the repository of all schemes of AD (external, conceptual, internal), as well as interrelations between the elements of these schemes representing the same and related concepts, has led to an additional source of contradictions. In accordance with the principle of unity of the basis of notion division, the following main characteristics can be distinguished for classification of inspections: - property of verification to be formally represented in the model's metascheme; - characteristic of the time of verification; - characteristic of the place of verification; - characteristic of the coverage of scheme elements. Each check, depending on its properties and characteristics, falls into one or another class for each of the criteria, and accordingly, appropriate means are applied for its implementation. As a result of the analysis of possible contradictions in the ERM scheme, a document containing 179 various checks was received. In order to be able to cancel complex modifications, in the storage systems, as a rule, transactions are applied. Generated by Oracle Designer (OD) API provides the ability to manage the start and end points of a transaction, which was used when creating the verification mechanism. The scheme changes that are accumulated during the transaction are checked as soon as possible in the presence of elements necessary for this. In the case of a negative result of the check, all data changes made during the transaction must be rollbacked, and the repository must accept the state it had at the time the transaction was opened. So, transaction model was adopted according to which each instance of the structural concept of the ERM model is added, modified or deleted within a single transaction. One of the important issues is the location of the verification mechanism in the system. The check should be performed when working in all tools that change the data stored in the repository. For each element and association type of ERM-objects there is its own procedure that realizes its verification. There are 3 blocks in each function: a block of mandatory checks, a block of checks for candidate objects and a block of checks of external objects. Thanks to these numerous versatile checks, the ERM repository ensures data scheme consistency. Prospects for the development of a mechanism of verifying ERM schemes for consistency are in the consistent use of principles of Model Driven Development. In accordance with them, it is necessary to declaratively present in the ERM metascheme all the conditions of consistency of schemes and implement universal program verifier that performs the necessary checks based only on this information. The semantic completeness of the ERM model is able to provide such descriptions.

Keywords

Authors

References