This page covers how to set up a local installation. If you are contributing to a repo that has enabled and you don’t want to run the hooks locally, all you need to do is push, observe if the CI passes, and if not, pull the auto-fixes made and then fix the remainder manually and push again. We recommend running the hooks locally too to avoid tedious feedback loops from CI for everything but trivial commits.


You can install the package from CRAN:


To access pre-commit functionality from R, you also need to install the pre-commit framework on which the hooks from this repo build. The following command line methods are tested to work with this R package (and accessing them from outside R is easier and use is slightly faster):

Alternatively, you can handle the installation from R using miniconda:

Then, in a fresh R session:

# once in every git repo either
# * after cloning a repo that already uses pre-commit or
# * if you want introduce pre-commit to this repo

The last command initializes pre-commit in your repo and performs some set-up tasks like creating the config file .pre-commit-config.yaml, where the hooks that will be run on git commit are specified. See ?precommit::use_precommit() to see how you can use a custom .pre-commit-config.yaml instead of the default at initialization. You can (obviously) change edit the file manually at any time.


The next time you run git commit, the hooks listed in your .pre-commit-config.yaml will get executed before the commit. The helper function precommit::open_config() let’s you open and edit the .pre-commit-config.yaml conveniently from the RStudio console. When any file is changed due to running a hook or the hook script errors, the commit will fail. You can inspect the changes introduced by the hook and if satisfied, you can add the changes made by the hook to the index with git add path/to/file and attempt to commit again. Some hooks change files, like the styler hook, so all you need to do to make the hook pass is git add the changes introduced by the hook. Other hooks, like the parsable-R hook, will need your action, e.g. add a missing closing brace to a call like library(styler, before they pass at the next attempt. If all hooks pass, the commit is made. You can also temporarily disable hooks. If you succeed, it should look like this:

See the hooks provided by this repo under vignette("available-hooks"). You can also add other hooks from other repos, by extending the .pre-commit-config.yaml file, e.g. like this:

-   repo:
    rev: v1.2.3
    -   id: check-added-large-files

To update the hook revisions, run precommit::autoupdate().


Do not abort while hooks are running in RStudio git tab. Non-staged changes are stashed to a temp directory and when you abort in RStudio, these changes are not brought back to you repo. Upvote this issue to change this. We hope that in the future, the changes will be recovered in RStudio too. Note that this is only an issue with RStudio. Stashes are restored when you abort a git commit with INT (e.g. Ctrl+C) on the command line. To restore stashes, manually after hitting abort in the RStudio git tab, you can git apply /path/to/patch_with_id whereas you find the patch under your pre-commit cache, which is usually under $HOME/.cache/pre-commit/.


If you used {precommit} before, upgrade these three components for maximal compatibility:


uninstall_precommit("repo") # just for the repo you are in.
uninstall_precommit("user") # remove the pre-commit conda executable.