Abstract:

It is now nearly 30 years since Peter Chen’s watershed paper “The Entity-Relationship Model – towards a Unified View of Data”. In his original paper Peter Chen looked at converting his new Entity Relationship (ER) model to the then existing data structure diagrams for the Network model which was the Database Management System (DBMS) which was prevalent at the time in industry. In recent years there has been a tendency to use Unified Modelling Language (UML) class diagrams for conceptual modelling for relational databases, and several popular course text books now use UML notation for the conceptual modeling stage of database design.

However Object and Relational technology are based on different paradigms. In the paper we argue that the UML class diagram is more of a logical model (implementation specific). ER Diagrams on the other hand, are at a conceptual level of database design dealing with the main items and their relationships and not with implementation specific detail. UML focuses on OOAD (Object Oriented Analysis and Design) and is navigational and program dependent whereas the relational model is set based and exhibits data independence. The ER model provides a well-established set of mapping rules for mapping to a relational model. The UML class model can be used as a conceptual model but there does not yet appear to be a well-established set of mapping rules.

An impedance mismatch is a problem that can arise from combining technologies from different paradigms. Previous work has identified 4 different levels of impedancy mismatch between Object and Relational technologies. Our work in this paper would fit into the paradigm impedance mismatch level and we describe a conceptual mismatch problem between ER diagrams and UML class diagrams.

In this paper we look specifically at concepts which can cause problems for the novice database designer due to this conceptual mismatch. Firstly, transferring the mapping of a weak entity from an Entity Relationship model to UML and secondly the representation of many-to-many relationships. We also look at the mixture of notations which students mistakenly use when modelling. This is often the result of different notations being used on different courses throughout their degree.

Several of the popular text books at the moment use either a variation of ER, UML, or both for teaching database modelling. Our paper also includes a survey of current popular text books and shows whether they use a variation of ER, UML, both ER and UML or a mixture of the two. At the moment if a student picks up a DBMS text book and looks at the database design modeling chapters they could be faced with either; one of the many ER variations, UML, UML and a variation of ER both covered separately, or UML and ER merged together.

Technology changes and modeling techniques and methods change, however when teaching the material to novice designers we should be aware of the conceptual impedance mismatch which exists between a UML class diagram and an ER diagram and the subsequent mapping to a relational database. There is a conceptual impedance mismatch and this will lead to confusion for students if they have previously studied conceptual modeling using an ER diagram.

Keywords:

UML class diagram, Entity Relationship modelling, impedance mismatch.