time • Enables automatically testing your package on winOS, macOS, linux, docker ◦ https://docs.github.com/en/actions/reference/virtual-environments-for-github-host ed-runners • Reproducible framework ◦ Particularly when you use docker http://bioconductor.org/help/docker/ ◦ powered by https://www.rocker-project.org/ • But it’s also cutting edge ◦ You might run into the latest R issues ◦ Might need to figure out how to install • https://github.com/r-lib/actions • https://www.jimhester.com/talk/2020-rsc-github-actions/
• Includes a GitHub Actions workflow that is Bioconductor-friendly ◦ https://lcolladotor.github.io/biocthis/reference/use_bioc_github_action.html ◦ Bioconductor docker + macOS + winOS ◦ R CMD build + check + BiocCheck ◦ covr: code coverage ◦ pkgdown: documentation website • Has functions similar to usethis but that are tailored for Bioconductor’s needs (and/or mine) ◦ https://lcolladotor.github.io/biocthis/reference/use_bioc_vignette.html creates a BiocStyle-based vignette with a structure that I like to use for my R/Bioconductor packages • Includes 4 dev/*.R scripts that will guide you throughout the process of making your R/BioC package
R build issues • Unlike BioC builds, you have to figure out how to deal with dependencies ◦ Upstream package might not be available • BioC 3.13 uses R-devel and CRAN has no binaries on macOS • Compiling errors can be harder to trace • Can use BioC 3.12 and/or focus on Bioconductor docker’s tests • Can be directly affected by rocker/BioC docker changes
Overall faster feedback than BioC’s builds ◦ Complementary, not a replacement • Can help notice issues with OS that you might not realize ◦ Example with tests failing on megadepth on macOS • Helped me be more productive this past round • dev/ scripts/guides can help new users become developers