https://github.com/NAIST-SE/18MillionLinksInCommitMessages/
This is a research artifact for the paper 18 Million Links in Commit Messages: Purpose, Evolution, and Decay. This artifact is a data repository including all 18,201,165 links associated with the information of languages and commit location (GitHub links including account names, repository names, commit hashes) and the processed data. The purpose of this artifact is enabling researchers to replicate our mixed-methods results of the paper, and to reuse our dataset of around 18 million links in commit messages for further research.
Since our full dataset exceeds the file size limitation of GitHub, we split some of the data into multiple files.
- Tools and data:
git-log-extractor
is the tool we used to collect the git logs, which where then further processed with Googlebigquery
.github-repo-list
contains the list of the GitHub projects in our stratified sample,git-clone-error
is the errors we observed while cloning some of the repos,git-clone-logs
is the output of our retrieval tool, andgit-commit-messages
the extracted data, i.e., the commit messages.gh-links
is the output of links to domain github.com pointing to same repository.
- Coding results: Target and purpose of links (n=1,145)
- Link evolution: Repeated links (n=46)
- Changes to link targets:
- Raw output of link target status comparison
- Link target and status according to Wayback Machine, retrieved using this script (n=1,145)
- Stack Overflow case study:
- We executed an SQL query to retrieve change events of Stack Overflow threads using the SOTorrent BigQuery dataset.
- Results
- Link decay: CSV files indicating whether links were dead (x) or alive (o) (n=6,667,207 distinct links)