ExceLint: automatically finding spreadsheet formula errors

OOPSLA 2018 |

DOI

Spreadsheets are one of the most widely used programming environments, and are widely deployed in domains like finance where errors can have catastrophic consequences. We present a static analysis specifically designed to find spreadsheet formula errors. Our analysis directly leverages the rectangular character of spreadsheets. It uses an information-theoretic approach to identify formulas that are especially surprising disruptions to nearby rectangular regions. We present ExceLint, an implementation of our static analysis for Microsoft Excel. We demonstrate that ExceLint is fast and effective: across a corpus of 70 spreadsheets, ExceLint takes a median of 8 seconds per spreadsheet, and it significantly outperforms the state of the art analysis.