Engaging developers in open source software projects: harnessing social and technical data mining to improve software development

TNM: A Tool for Mining of Socio-Technical Data from Git Repositories

“…Methods to infer socio-technical data from Git repositories include applying the PageRank algorithm to evaluate the influence of a commit on other commits [19], computing the degree of knowledge to evaluate the properties of developers' contributions [12], [13], analyzing collaboration in a project to evaluate its risk of stalling [6], [15]- [17], determining code ownership and code authorship to find code components and source files at risk of having quality issues [3], [14], and using socio-technical congruence to measure the alignment between the coordination required by the project technical dependencies and the actual coordination between its members [24], [29].…”

“…Degree of interest part is a measure of developers' interactions with code; it is necessary to track developers' behavior to calculate it. Some of the later implementations of the degree of knowledge rely only on data from Git [12], [13]. The obtained degree of knowledge can be then used to study individual contribution patterns [13] or to compute the bus factor of the project [15] (see Section II-C4).…”

“…• FilesOwnershipMiner is based on the Degree of Knowledge (DOK) [12], which measures the knowledge of a developer or set of developers about a particular section of code. The miner yields a knowledge score for every developer + file pair in the form of a nested map, and information about the authorship of code on a line level.…”

“…A number of prior studies investigated collaboration between engineers using information extracted from VCS repositories. Examples include computing the degree of knowledge to study developers' contributions [12], [13], analyzing code ownership to identify bug-prone components [3], [14], estimating existential risks of projects [6], [15]- [17], and applying the PageRank [18] algorithm to infer causal relationships between the commits [19].…”

“…TNM incorporates implementations of several established data mining techniques, or individual miners. Namely, it includes a miner for the degree of knowledge-based files ownership [12], a miner applying the PageRank algorithm [18] to the analysis of a commit history for commit influence evaluation [19], miners extracting data for the analysis of socio-technical congruence [24], a miner for calculating the distribution of commit time over the week, and one for finding files that have been edited by several developers. Moreover, TNM can be easily extended to include other analysis algorithms.…”