Multi-extract and Multi-level Dataset of Mozilla Issue Tracking History

Abstract

Many studies analyze issue tracking repositories to understand and support software development. To facilitate the analyses, we share a Mozilla issue tracking dataset covering a 15-year history. The dataset includes three extracts and multiple levels for each extract. The three extracts were retrieved through two channels, a front-end (web user interface (UI)), and a back-end (official database dump) of Mozilla Bugzilla at three different times. The variations (dynamics) among extracts provide space for researchers to reproduce and validate their studies, while revealing potential opportunities for studies that otherwise could not be conducted. We provide different data levels for each extract ranging from raw data to standardized data as well as to the calculated data level for targeting specific research questions. Data retrieving and processing scripts related to each data level are offered too. By employing the multi-level structure, analysts can more efficiently start an inquiry from the standardized level and easily trace the data chain when necessary (e.g., to verify if a phenomenon reflected by the data is an actual event). We applied this dataset to several published studies and intend to expand the multi-level and multi-extract feature to other software engineering datasets.