Qualitative deviation models vs. quantitative models for fault localization in spreadsheets

Automatizing fault localization in programs has been an interesting and active research field for several decades. In order to increase efficiency at least tool support for debugging would be highly required. This holds even more for end-user programs like spreadsheets. In this paper, we discuss three different models for spreadsheet debugging where one relies on qualitative algebra and deviations between the computed and the expected values of spreadsheet cells. The main objective of this paper is to clarify the questions whether a qualitative model for debugging can replace a model capturing the semantics of a spreadsheet. In order to answer this question, we carried out an empirical evaluation based on a publicly available spreadsheet corpus showing that qualitative models allow for computing similar diagnoses but require a fraction of runtime. Hence, qualitative models have the potential of being used for fault localization under real-time conditions.