On the Utility of Adding An Abstract Domain and Attribute Paths to SQL

Abstract

Albeit its popularity today, RDBMS and the relational model still have many limitations. For example, one needs to pay premature attention to naming issues in the schema designing phase; and the syntax for conjunctive queries is verbose and redundant, especially for multi-table joins and composite primary/foreign keys. In this thesis, we introduce and explain the method to handle and resolve these issues that is proposed by Borgida, Toman, and Weddell: the conceptual schema that supports abstract relations and attributes, and an extended query language SQLpath built on top of standard SQL that supports the usage of attribute paths and abstract attributes in queries. We demonstrate a systematic approach to map a database schema expressed in the relational model to the abstract relational model and illustrate how to write SQLpath queries with attribute paths to solve query problems involving complex table joins. This thesis can serve as both an introduction and tutorial to abstract database modelling and the SQLpath query language.

Additionally, we performed an empirical experiment to evaluate the performance of SQLpath when solving real database query problems by employing students with prior experience with SQL to read and write SQLpath queries and recorded their accuracy and time consumption against usage of regular SQL. The result of this experiment is presented in this thesis, including a statistical analysis of the results. In short, we uncover evidence that SQLpath is more efficient to use for both reading and writing conjunctive and alike queries, especially for non-trivial cases where multiple constraints were required. However, while SQLpath can hide explicit table joins when writing queries spanning multiple intermediate tables, whether this benefit can make users produce more accurate results still remains unclear as we were not able to draw any conclusion from collected data due to lack of statistical significance.