In general, we recommend following the Tidyverse style guide as much as possible. There are two R packages that support this style guide out of the box:
Both these tools can be customised to disable certain checks or using different values. However, we recommend sticking to the default values as much as possible.
For VSCode users, the vscode-R plugin provides most of these features in VSCode as well.
A note on the use of the pipe (
|>or%>%) operator, although very useful for data analysis code, we recommend to avoid their usage in R package code. The main reason is that pipes make code more difficult to debug, especially for very long sequences of pipes, as the intermediate objects are not immediately accessible. In addition, the use of%>%would incur an additional dependency.
The R Packages book provides a comprehensive guide to R package development, which closely follows the Tidyverse style guide (as they’re from the same authors). We recommend following the best practices described in this book.
To help with adhering to these practices and make your life as an R developer much easier, there are the
devtools and usethis packages. From setting up your
package file structure with usethis::create_package(), running unit tests and other checks with devtools::check(),
to publishing your package to CRAN with devtools::release(), these two packages have it all covered. In fact, whenever
you need to do anything other than edit an *.R file, there is probably a
usethis function function you can (and should!) use instead. And if
there is not, ask yourself whether you really need to do that thing.
And of course, both these packages play nicely together with RStudio, with several built-in features and shortcuts (see the R Packages book for tips and tricks).
The r-lib/actions repository contains several reusable GitHub Actions (GHA)
workfklows for R projects (not only packages). The examples
directory provides complete workflows for common R project CI tasks. All of these can be easily set up using
usethis::use_github_action().
When developing an R package for the Bioconductor project, a few exceptions to the above recommendations should be considered. Bioconductor provides their own list of recommendations. These are largely similar to the Tidyverse style guide, with a few exceptions:
camelCase for function namesThe reason for the latter is that Bioconductor packages make extensive use of the S4 OOP
framework in R. Within this framework, the convention is to use CamelCase for Class names and camelCase for method
names.
The most important of these rules are checked by
BiocCheck::BiocCheck() The
biocthis package provides a handy alternative to
usethis for Bioconductor packages. Finally, there is a bioc-actions with
Bioconductor-specific GHA workflows.
The biocthis package provides
biocthis::bioc_style() to configure the syler
package with the Bioconductor-specific rules. You can set
options(styler.addins_style_transformer = "biocthis::bioc_style()") in your ~/.Rprofile to configure these rules
globally, or in a project-specific .Rprofile.