The thesis is conducted within the field of software testing and was motivated by test challenges at The Norwegian Tax Department. It focuses on test automation in the context of functional black-box regression testing of database applications, based on model specifications from classification tree models. The main contributions of the thesis are:
- A novel and practical approach to automated, system-level regression testing of database applications, that uses classification tree models for black-box test specifications.
- An industrial case study of the proposed regression test approach, by applying a novel tool implementation, DART ([DA]tabase [R]egression [T]esting), to the business critical batch jobs of a large database application in a public-service setting.
- A thorough investigation of strategies for selecting test cases generated from classification tree models, evaluated in the context of black-box regression testing of database applications.
- The definition and evaluation of a clustering strategy for grouping regression test deviations according to their root causes in order to help scale their inspection and analysis.
- A practical and novel approach for matching production data against classification tree models in order to (1) detect model coverage, and (2) reduce the level of redundancy, and thus test effort, by selecting a subset of test cases for test execution.
- A practical approach for automatic test data generation based on classification tree models.
- An assessment of the combination of production data and generated test data for regression testing, following various combinatorial test strategies based on classification tree models.
(In the list above, the term “practical” means scalable to large systems, while relying on information that can realistically be provided by test engineers or domain experts.)
Quotes from the adjudication committee’s evaluation of the thesis:
- ”The thesis is an excellent example of research that supports evidence-based practice.”
- “The work is unique in its success to conduct the research in a real-world, large-scale context of an operational database system, and yet deliver both high-quality scientific research and actionable recommendations for the practitioners.”
- “The committee finds the thesis making a significant contribution to the research field of empirical software engineering, and in particularly a distinct contribution to the regression testing of large database applications.”
The work has been conducted at Simula Research Laboratory in collaboration with the Norwegian Tax Department.
Prior to the defense, at 10:00, Erik Rogstad will present his trial lecture The Role of Case Studies, Technical Action Research and Design Science within Research on Regression Test Processes.
The adjudication committee
- Associate professor Amela Karahasanovic, University of Oslo
- Professor Per Runeson, Lund University, Sweden
- Professor Mika Mäntylä, University of Oulu, Finland
Chair of the disputation
Lecturer Dag Langmyhr, University of Oslo
- Professor Lionel Briand, University of Luxembourg
- Chief research scientist Arnaud Gotlieb, Simula Research Laboratory