Analyzing Large Spreadsheet Programs

Although the results of spreadsheet programs are the base for very important decisions and are subject to many changes, they are only poorly documented. In this paper we introduce an approach that extents previous work in the area of spreadsheet visualization. It enables the maintainer to decompose large spreadsheets into self contained parts, that are similar to modules in conventional software. Of course there are important differences, as these modules are only recognized during analysis, and users are not forced to build modular spreadsheets. This is very important, as we aim not to change the spreadsheet users. It has turned out, that attempts to change the users are usually doomed to fail. This approach can be considered orthogonal to semantic classes, that have been introduced in prior work. The generated abstraction is not build upon the formulacontents of the spreadsheet, but on properties of the data flow graph. Therefore, spreadsheets with irregular formulas can be solely analyzed with the new approach. On the other hand, large spreadsheets can be decomposed into data modules at first, that are smaller and easier to understand. Each of the data modules can than be analyszed with a formula based approach.