Skip to content

Repositories

Most R packages are developed in a public repository. Here is where those repositories sit, what they carry, and how long work stays open.

16,803
Packages with a repository
16,654
Distinct repositories
43
Hosts represented

Hosting

Where R package repositories are hosted, and how the share breaks down.

github.com 16,358 97.4%
gitlab.com 153 0.9%
codeberg.org 94 0.6%
r-forge.r-project.org 67 0.4%
bitbucket.org 31 0.2%
forge.inrae.fr 17 0.1%
37 other hosts 83 0.5%

Out of 16,803 packages with a repository we could read.

Repository contents

Scanned from the repository tree. A package may follow a practice by other means, so these counts are lower bounds.

Continuous integration

Automated testing and build services configured in the repository.

Checks on R-devel is a share of the repositories with an R CMD check workflow, not of every repository scanned.

Any CI configured, Travis config alone not counted 10,585 65.6%
GitHub Actions 10,348 63.7%
R CMD check workflow 8,636 53.5%
Coverage workflow 4,112 25.5%
Checks on R-devel 7,033 81.4%
GitLab CI 98 0.6%
Travis CI (config file) 1,813 11.2%
AppVeyor 525 3.2%
CircleCI 44 0.3%

Review and style

Tools and conventions that shape how code is written and reviewed.

A pull request template counts when GitHub shows one for the repository, including an organization default from its .github repository.

From Sep 26, 2026 the scan also finds pull request templates under lower-case names, in more folders and as organization defaults, so that count rose that week.

lintr 1,033 6.4%
Lint workflow 778 4.8%
jarl 47 0.3%
Air 574 3.5%
pre-commit 230 1.4%
EditorConfig 188 1.2%
Code owners 168 1.0%
Pull request template 588 3.6%

Reproducibility

Infrastructure for recreating the development or analysis environment.

renv 343 2.1%
Dockerfile 199 1.2%
Dev container 148 0.9%
Nix 29 0.2%
Binder 18 0.1%
data-raw 2,908 17.9%

Documentation

Package websites and how they are deployed.

pkgdown site 7,248 44.7%
altdoc site 17 0.1%
litedown site 8 <0.1%
Quarto project 37 0.2%
GitHub Pages site 8,374 51.9%

For agents and models

Files published for machine readers rather than people.

Provides agent skills 17 0.1%
llms.txt 35 0.2%

Credit and governance

How the project tracks contributions, funding, and decision-making.

A code of conduct or contributing guide counts when GitHub shows one for the repository, including an organization default from its .github repository.

Code of conduct 2,350 14.6%
Contributing guide 1,947 12.1%
CITATION.cff 721 4.4%
codemeta 568 3.5%
Zenodo 121 0.7%
JOSS 107 0.7%
All Contributors 8 <0.1%
Governance 10 0.1%
Security policy 166 1.0%
Funding 249 1.5%

Release process

How the default branch compares with the release on CRAN.

The share is of the repositories whose default branch version could be compared with the CRAN release, not of every repository scanned.

Default branch ahead of CRAN 4,426 31.8%

Out of 16,242 repositories scanned. Detection methodology

Time to close

Each repository contributes its own median. Only those with at least one closure are included.

Days to close an issue

The repository median, so half of its closed issues took longer.

up to 1 day 1,691 16.7%
1 day to 7 days 2,492 24.6%
7 days to 1 month 2,587 25.6%
1 month to 3 months 1,580 15.6%
3 months to 1 year 1,212 12.0%
over 1 year 552 5.5%

10,114 repositories have this measurement.

Days to close a pull request

Closed covers merged and rejected alike.

up to 1 day 7,254 74.7%
1 day to 7 days 1,272 13.1%
7 days to 1 month 582 6.0%
1 month to 3 months 267 2.7%
3 months to 1 year 228 2.3%
over 1 year 112 1.2%

9,715 repositories have this measurement.

Notes

How repositories are counted

Repository URLs come from each package's DESCRIPTION file. A repository serving multiple packages counts once in the distinct total but once per package elsewhere. 75 of these repositories are marked archived by their host while the package is still published. Practice counts are lower bounds: the weekly scan sees only what is committed, and anything in .Rbuildignore is invisible.

Privacy choices

These apply to this browser and are stored on this device only. Nothing about your choice is sent to us.

Read the privacy policy