Methods: Analysis

We produced several datasets from the data we collected from ePrints repositories and GitHub, as well as visualisations.

Repository mention intent

We manually examined all publications with links in the first two pages to find whether these repositories were created for the publication or simply used/cited. The resulting dataset, eprints_w_intent.csv, contains a row for each repository-publication pair examined in this manner:

column name	type	description	comment
github_repo_id	str	GitHub repository ID	<username>/<reponame>
mention_created	bool	whether the repository was mentioned in the publication as having been created for it
pub_title	str	title of the publication
pub_author_for_reference	str	one of the publication authors as listed on ePrints	not reflective of main authorship
pdf_url	str	link to downloadable file for the publication
page_no	int	number of the page the GitHub link was found on	first page denoted by 0
detected_github_url	str	GitHub link as extracted from the publication during PDF parsing
pattern_matched_github_url	cleaned GitHub link from regular expression matching
eprints_date	str	date field in ePrints	YYYY-MM-DD
eprints_pub_year	int	year from the date field
eprints_repo	str	identifier of the ePrints repository the publication was found in

Aggregated timelines

We merged all repository metadata into one file with one row per repository:

aggregated_overall.csv:

column name	type	description	comment
github_user_cleaned_url	str	GitHub repository ID
archived	bool	whether the repository was archived
created_at	datetime	time of repository creation
has_wiki	bool	whether the repository has a wiki
has_pages	bool	whether the repository has GitHub pages
license	str	license identifier	as provided by GitHub
readme_size	int	README size in Bytes
readme_path	str	path to README file
readme_emojis	int	number of emojis detected in the README file
contributing_size	int	size of CONTRIBUTING.md in bytes	0 if non-existent
citation_added	datetime	time of commit adding CITATION.cff
contributing_added	datetime	time of commit adding CONTRIBUTING.md
repo_created_at	datetime	same as created_at	leftover duplication from dataframe merge
week_since_repo_creation_citation_added	int	number of weeks between repository creation and addition of CITATION.cff
week_since_repo_creation_contributing_added	int	number of weeks between repository creation and addition of CONTRIBUTING.md
license_type	str	one of permissive, unknown, none or non-permissive	based on https://en.wikipedia.org/wiki/Permissive_software_license
readme_size_class	str	one of "none", "ultra-short", "short", "informative", "detailed"	empirically determined: between 1 and 300 bytes is ultra-short, 300-1500 is short, anything between 1500 and 10000 is informative, everything beyond is detailed
forks_count	int	number of forks of the repository
stars_count	int	number of stars of the repository
max_active_contributors	int	maximum observed number of active contributors to the repository in any week	an active contributor has made at least one commit in the last 12 weeks

Moreover, we reframed the data collected from GitHub as timelines - one row for each week of repository life. The resulting data files are:

aggregated_timeline.csv:

column name	type	description	comment
github_user_cleaned_url	str	GitHub repository ID
week_since_repo_creation	int	week since repository creation	increasing with repo creation being week 0
closed_count	int	number of closed issues for this repository in that week	total number of closed issues up to and in that week, not the number of issues closed in that week
open_count	int	number of open issues in that week	as for closed_count
active_contributors	int	number of active contributors in that week	a contributor is considered active if they have made at least one commit to the main branch in the last 12 weeks
contributors	int	number of users who have ever made a commit to the main branch between repository creation and that week
forks_count	int	number of forks up until and in that week
stars_count	int	number of stars up until and in that week
ownership_added	bool	whether an ownership keyword was added to the README file in that week	addition means that the line showed up in commit additions
usage_added	bool	whether a usage keyword was added to the README file in that week	as for ownership_added
citation_added	bool	whether a citation keyword was added to the README file in that week	as for ownership_added
citation_file_added	bool	whether a CITATION.cff file was added in that week
contributing_file_added	bool	whether a CONTRIBUTING.md file was added
paper_published	bool	whether a publication citing this repository was published that week	based on ePrints date field with potential limitations as outlined elsewhere

aggregated_commit_author_timeline.csv:

column name	type	description	comment
github_user_cleaned_url	str	GitHub repository ID
author	str	username of a contributor to the repository
week_since_repo_creation	int	week since repository creation	increasing with repo creation being week 0
commits	int	number of commits made by author in week_since_repo_creation
active_contributors	bool	whether author is considered an active contributor in this week	a contributor is considered active if they have made at least one commit to the main branch in the last 12 weeks

aggregated_issue_user_timeline.csv:

column name	type	description	comment
github_user_cleaned_url	str	GitHub repository ID
week_since_repo_creation	int	week since repository creation	increasing with repo creation being week 0
user	str	username of user who interacts with the repository's issues
created_count	int	number of issues created by user in week_since_repo_creation in the repository
closed_count	int	number of issues closed by user in week_since_repo_creation in the repository
user_status	str	one of [inactive', 'opening', 'both', 'closing']	opening if at least one issue opened in last 12 weeks, closing if at least one issue closed in last 12 weeks, both if both apply, inactive otherwise