On the Diversity of Software Package Popularity Metrics: An Empirical Study of npm
Download as PPTX, PDF0 likes44 views
This document summarizes a study on the diversity of software package popularity metrics using data from the npm registry. The study analyzed 9 different metrics for 175,774 packages to determine how correlated the metrics were. It found the metrics were not strongly correlated, implying that using different metrics could produce different outcomes. It concludes there is a need to define a framework that accounts for the diversity and context-dependence of software popularity metrics.
On the Diversity of Software Package Popularity Metrics: An Empirical Study of npm
- 1. On the Diversity of Software Package Popularity Metrics: An Empirical Study of npm Ahmed Zerouali, Tom Mens, Gregorio Robles, Jesus Gonzalez Barahona The IEEE International Conference on Software Analysis, Evolution and Reengineering (SANER) Hangzhou, China - February 24-27, 2019
- 2. /Motivation
- 3. /Motivation
- 4. /Goal How are metrics of package popularity related to each other?
- 6. /Method/ 9 popularity metrics # runtime dependent repositories (libraries.io) # transitive runtime dependents (libraries.io) # direct runtime dependents (npm and libraries.io) # downloads (npm) # npm stars (npm) # github stars (GitHub) # forks (GitHub) # pull requests (GitHub) # subscribers (GitHub)
- 7. /Metrics Emanating From the Same Source
- 8. /Metrics Emanating From the Same Source
- 9. /Metrics Emanating From the Same Source
- 10. /Metrics Emanating from Different Sources/ 6 metrics # runtime dependent repositories (libraries.io) # direct runtime dependents (npm and libraries.io) # downloads (npm) # npm stars (npm) Aggarwal-Popularity = #forks + #stars + #pull requests² (GitHub) # subscribers (GitHub)
- 11. /Metrics Emanating from Different Sources/ 6 metrics
- 12. /Metrics Emanating from Different Sources/ 6 metrics
- 13. /Limitations - There are other metrics not considered in the study (e.g. development dependents, SoureRank (from libraries.io), etc) - We focused only on packages that are hosted on GitHub. - We focused only on one package repository, npm.
- 14. /Conclusion - Conducted analysis on the diversity of software popularity metrics. - Popularity metrics are not strongly correlated, implying that the use of different metrics may produce different outcomes. - There is a need to define a measurement framework that takes into account the diversity and context- dependence of software popularity.
- 15. /Future Work - Do a more fine grained survey on software popularity metrics. - Consider concrete examples of studies and reproduce them using other popularity metrics. - Extend the analysis on other ecosystems.