Automated Repair of Data Faults in Templated Spreadsheets

Spreadsheets are widely used in data recording and data analysis in various fields. Since most spreadsheet users are non-programmers, they are prone to various errors in the process of designing and using the spreadsheet, which may bring huge economic losses to the organization. Therefore, a lot of related work has emerged to support the users in detecting and fixing errors in spreadsheets. In these studies, most of them are based on the assumption that the original data is error-free, because formulas in the spreadsheet are considered more error-prone than input cells. However, the assumption is not reasonable in the field of frequently using spreadsheet templates, because they have been professionally validated during usage. In such cases, root errors often arise from wrong input values rather than cells with formulas. Therefore, this paper proposes an error repair method for input cells in templated spreadsheet based on symbol execution and constraint satisfaction solving. First, find the output cells in the spreadsheet according to dependencies. Second, for suspicious output cells, the input cells that determine their results are obtained via symbolic execution. Finally, the repair candidates of the input cells are obtained by constraint satisfaction solving with domain knowledge.