In lorenzwalthert/precommit: Pre-Commit Hooks

Below is a comprehensive list with all hooks from {precommit} as well as their arguments or flags, if they take any. With the standard config file that gets placed in your project with precommit::use_precommit() all hooks should work out of the box, but you can further customize them as described below. Other repos also host hooks, many are listed here.

Good to know

File modification

Some hooks will fail without changing the files you want to commit, like the lintr hook - and you need to make manual changes for the hook to pass on the next attempt. Other hooks like the roxygenize hook write to files, and if that changes the file, the hook will fail, but this means for most hooks you won't need to modify the file manually after the attempted commit, just stage the changes and try to commit again. Below, we indicate for every hook if it modifies files or not.

Arguments

Arguments for the hooks are specified as described in the pre-commit.com documentation.[^1] You can specify arguments like this:

[^1]: Note that there might be issues with arguments that contain special characters and you might have to quote them and the order of single, double and escaped quotes may not give identical results on one platform and may not be portable to all platforms either.

library(magrittr)
rev <- paste0("v", packageVersion("precommit"))
cat(glue::glue("
repos:
-   repo: https://github.com/lorenzwalthert/precommit
    rev: {rev}
    hooks:
    -   id: lintr
        args: [--warn_only, --key=value]
"))

Other hook settings

Apart from specifying the args key as described above, there are other hooks settings you can specify, e.g. file exclusion. If you don't, they are inherited from the default repository's configuration (i.e. the .pre-commit-hooks.yaml file in https://github.com/lorenzwalthert/precommit). See the pre-commit documentation, for the available settings.

Hooks

style-files

A hook to style files with styler. Only commit code corresponding to the tidyverse style guide. You can pass arguments to style_file(...) using the --key=value syntax like this:

  id: style-files
  args: [--scope=spaces, --reindention=specify_reindention('#')]

In addition, the hook takes the following arguments that are not passed to style_file(...):

• Argument style_pkg and style_fun default to styler and tidyverse_style. If you want to use another style guide than the tidyverse style guide, you can specify it like this:

  id: style-files
  args: [--style_pkg=pkg.with.style.guide, --style_fun=exported.style.function]

• The argument --no-warn-cache is deprecated and will be removed in a future release. Please remove it from your .pre-commit-config.yaml.

• Argument cache-root is passed to options() to set styler.cache_root. Default value: styler-perm. The argument determines the sub-directory under the {R.cache} cache directory that {styler} uses. If you want {styler} to auto-clean up caches older than 6 days, set this to "styler". For more information, see help("caching", package = "styler").

  id: style-files
  args: [--cache-root=styler]

• Argument ignore-start and ignore-stop is passed to options() to set styler.ignore_start and styler.ignore_stop respectively. Not set by default, so the {styler} defaults apply. This was introduced in {precommit} 0.2.2.9012. For example, if you want to restore old behavior of styler \< 1.6.2, where only the literals styler: off and styler: on were accepted, use this regex:

  id: style-files
  args: [--ignore-start="^# styler: on$", --ignore-stop="^# styler: off$"]

This hook modifies files unless you specify --dry=fail (requires {styler} > 1.3.2).

readme-rmd-rendered

Make sure README.Rmd hasn't been edited more recently than README.md, i.e. remind you to render the .Rmd to .md before committing.

This hook does not modify files.

parsable-R

Checks if your .R and .Rmd files are "valid" R code by checking if running parse() on them (or their knitr::purl()ed output for .Rmd) returns an error.

This hook does not modify files.

parsable-roxygen

Checks if roxygen comments within your .R files are "valid" by checking if running roxygen2::parse_file() on them returns any messages.

This hook does not modify files.

eval

By default, each file will be parsed, but code will not be evaluated - neither any explicit code in the file, nor any @eval tags within roxygen comments.

If your commentary contains @eval tags which you would prefer to evaluate, you can specify the --eval flag, which will cause the file's code to be evaluated in an environment created by roxygen2::env_file(). Note that dependencies of the code to evaluate must be available for pre-commit. You may list these as additional_dependencies: for the parsable-roxygen hook in .pre-commit-config.yaml.

Inline R code within roxygen comments (i.e. within backticks) is not evaluated by this hook, whether or not --eval is specified. You would need to run the roxygenize hook for that.

  id: parsable-roxygen
  args: [--eval]

This hook was added in version 0.4.3.9000.

no-browser-statement

Guarantees you that you don't accidentally commit code with a browser() statement in it.

This hook does not modify files.

no-print-statement

Guarantees you that you don't accidentally commit code with a print() statement in it.

This hook does not modify files. This hook was added in version 0.3.2.9020.

no-debug-statement

Guarantees you that you don't accidentally commit code with a debug() or debugonce() statement in it.

This hook does not modify files. This hook was added in version 0.2.2.9012.

spell-check

Checks spelling with spelling::spell_check_files().

Excluded files

When you invoke precommit::use_precommit() and .pre-commit-config.yaml is written to your repo (unless you specify config_source otherwise), we copy the expression in the exclude: key from spell check hook the default repository's configuration (i.e. the .pre-commit-hooks.yaml file in https://github.com/lorenzwalthert/precommit) into your config file, so you can easily add or remove some files. As of r paste0("v", packageVersion("precommit")), the following regex is used to exclude files following the verbose python regex syntax:

readLines(system.file("pre-commit-hooks.yaml", package = "precommit")) %>%
  gsub("^ *exclude *: *>", "    exclude: |", .) %>%
  yaml::yaml.load() %>%
  purrr::keep(~ .x$id == "spell-check") %>%
  magrittr::extract2(1) %>%
  magrittr::extract2("exclude") %>%
  cat(sep = "\n")

language

The lang arg will be passed to spelling::spell_check_files().

  id: spell-check
  args: [--lang=<language>]

read only

The --read-only flag will be passed to spell check. This flag makes this hook idempotent.

  id: spell-check
  args: [--read-only]

This hook does not modify input files. It will add all words not found in the dictionary to inst/WORDLIST, assuming they were spelled correctly but were not in the dictionary. An example might be "RStudio". The hook error message will contain all words written to inst/WORDLIST, so if there were really some typos, make sure to fix them and remove them from inst/WORDLIST. If there were not typos, or you fixed all, stage inst/WORDLIST and this time, the commit should pass.

To opt out of updating inst/WORDLIST provide the --read-only flag.

roxygenize

A hook to run roxygen2::roxygenize(). Makes sure you commit your .Rd changes with the source changes. To take advantage of caching, you don't need to run roxygen2::roxygenize() manually anymore. The argument --no-warn-cache is deprecated and will be removed in a future release. Please remove it from your .pre-commit-config.yaml.

Because the hook will write the version of {roxygen2} to DESCRIPTON, you should either make sure the version you use when you call {roxygen2} interactively matches the one from in {precommit} or simply not run {roxygen2} manually.

If you specify additional roclets through the Roxygen: field in DESCRIPTION, e.g. from {pkgapi} you must specify the dependencies explicitly such that renv::install() understands it, e.g.

  id: roxygenize
  additional_dependencies:
  - r-lib/pkgapi

This hook does not modify input files, but writes to .Rd files in man/, NAMESPACE and potentially others depending on which roxygen roclets you specified in DESCRIPTION.

Arguments

id: roxygenize
args: [--root=<R package root>]

• Argument root specifies the directory in the git repo that contains the R package. Defaults to . since for most R package git repos, the git and R package root coincide. Added in version 0.3.3.00000.

deps-in-desc

Checks if packages used with the pkgname::fun() syntax are listed in your DESCRIPTION file. Note that README.Rmd is never checked.

Arguments

• Flag allow_private_imports lets the user specify that private imports into the package namespace are tolerable, e.g. somepkg:::x(). Flag not set by default, i.e. the hook will fail if such a call is found.

  id: deps-in-desc
  args: [--allow_private_imports]

• Argument root specifies the directory in the git repo that contains the R package. Defaults to . since for most R package git repos, the git and R package root coincide. Added in version 0.3.2.9000.

id: deps-in-desc
args: [--root=<R package root>]

This hook does not modify the file DESCRIPTION because the user should decide for each package if it should go to Imports: or Suggests:, which can be done easily with usethis::use_package().

use-tidy-description

A hook to run usethis::use_tidy_description() to ensure dependencies are ordered alphabetically and fields are in standard order.

This hook does modify the file DESCRIPTION.

Arguments

id: use-tidy-description
args: [--root=<R package root>]

• Argument root specifies the directory in the git repo that contains the R package. Defaults to . since for most R package git repos, the git and R package root coincide. Added in version 0.3.3.00000.

lintr

A hook to run lintr::lint() to check that R files are lint free. Argument warning_only changes the behavior of the pre-commit to be non-blocking. You should set this with the field verbose: true.

  id: lintr
  args: [--warn_only]
  verbose: true

When configured this way, lintr prints lint errors as they appear. Other arguments are not supported. Instead, lintr config should be specified in a .lintr config file in Debian Control Field Format as specified in the .lintr documentation.

This hook does not modify any file.

codemeta-description-updated

Make sure DESCRIPTION hasn't been edited more recently than codemeta.json,

i.e. remind you to run codemetar::write_codemeta() in order to keep codemeta.json in sync with DESCRIPTION.

This hook does not modify any file.

Arguments

id: codemeta-description-updated
args: [--root=<R package root>]

• Argument root specifies the directory in the git repo that contains the R package. Defaults to . since for most R package git repos, the git and R package root coincide. Added in version 0.3.3.00000.

pkgdown

Check if your {pkgdown} config file (e.g. _pkgdown.yml in your root) has the correct entries for references and articles. This hook skips the time-consuming parts of building the index and reference and only performs the validation. Hence we don't rely on the extensive dependency graph of {pkgdown} being installed, including packages with heavy build-time dependencies and system libraries.

For this check, we rely on the the global R package library and require all development dependencies of the package you want to run this hook for to be installed, as well as {pkgdown} (without its dependencies).

This hook does not modify files. Added in version 0.3.2.9003.

renv-lockfile-validate

Guarantees that you don't accidentally commit an invalid renv.lock file. See renv::lockfile_validate() documentation for details.

The below config that uses only --error should suffice for most users.

id: renv-lockfile-validate
args: [--error]

This hook does not modify files. Added in version 0.4.3.9005.

Arguments

id: renv-lockfile-validate
args: [--schema=<schema>] [--greedy --error --verbose --strict]

lorenzwalthert/precommit documentation built on Feb. 7, 2025, 4:31 p.m.