Data modeling using XML schemas.

摘要

XML has established itself as a powerful and widely used logical data model. This is because of several reasons (a) XML is the standard for information exchange over the World Wide Web; therefore, if multiple applications want to communicate with each other over the web, they need to represent their data in XML, (b) XML provides several convenient features for data modeling, such as representing relationships using hierarchies, and IDREF, which allows us to navigate relationships using path expressions, rather than joins, and (c) XML provides several new and useful features for data modeling, needed for modern day applications, such as union types, and recursive types; it is also possible to represent ordered relationships using XML.; A data model can be considered as consisting of three aspects: structural specification, constraint specification, and operations. A schema (or data definition) language describes how structures and constraints can be specified. Different schema languages have been proposed for XML and are being evaluated, these schema languages have different characteristics with respect to structural and constraint specification. Our work tries to study what features are required in XML schema languages for database applications. We study this by considering conceptual models like Entity Relationship (ER) model, extended with category concept and the ability to specify ordered relationships, as a fair representative of database application requirements. We call our conceptual model for XML as ERex (ER extended for XML). By studying how ERex models can be represented as XML, we are able to examine the different features needed in XML schema languages. We conclude (a) structures should be specified using “single type tree grammars”, rather than other subclasses of regular tree grammars, and (b) constraints should be specified using “types”, rather than “path expressions”.; We further study different issues with respect to data modeling with XML: (a) We define different normal forms for XML, (b) We compare two subtyping schemes for XML: “explicit subtyping” and “implicit subtyping”, (c) Based on our study for representing ERex models in XML and vice versa, we study how relational representations can be translated to XML, and how XML representations can be translated to relational representations.