Jarl, a new R linter

A linter?


Tool to find code that is not optimal, likely wrong, hard to read, etc.


It performs static analysis.


Most languages have at least one linter.


A linter is opinionated.

What do you mean opinionated?


x = y * .52
if (all.equal(x, z)) {
  print("x and z are equal")
}

All possible violations:

  • x = y * .52

    • not important: use <- instead of =

    • not important: use 0.52 instead of .52

  • important: all.equal(x, z) doesn’t return FALSE if x != z

Current status


So far in R: lintr

  • pioneer for linting in R
  • 117 rules (formatting + linting), e.g. any(is.na(x)) -> anyNA(x)

But:

  • reports lints, does not fix them
  • can be slow
  • requires an R setup

Introducing Jarl


  • Released in November 2025
  • ~ 85 rules (linting only)
  • fast (written in Rust)
  • automatic fixes when possible
  • single binary to run
  • editor integration

Plan


  1. How to use


  1. How to configure


  1. How to contribute

1. How to use

1. How to use


Three ways to use Jarl:

  1. in terminal
  2. with extensions in your editor
  3. in CI

1. How to use in terminal


Basic usage:

# Run linter on current directory
jarl check .


Recommended usage: run this first

jarl check . --statistics

and then proceed rule by rule (and possibly tweak the configuration).

1. How to use in terminal


Rules can come with safe fixes, unsafe fixes, or no fixes at all:

# Apply safe fixes only
jarl check . --fix

# Apply safe and unsafe fixes
jarl check . --fix --unsafe-fixes

Note

By default, fixes are not applied if:

  1. the project doesn’t use version control
  2. or the version control status is “dirty” (some files are unstaged)

Use --allow-no-vcs and --allow-dirty carefully!

1. How to use with extensions



Install the “Jarl” extension in your editor (Positron, Zed, VS Code, Neovim, Helix, etc.). RStudio integration is not possible.


Using the extension enables code highlighting and code actions.

1. How to use in CI


Use the GitHub actions workflow: setup-jarl

Silencing Jarl


For the occasional false positive, use suppression comments:

# jarl-ignore <rule name>: <reason for skipping>
# jarl-ignore any_is_na: this is a demo
any(is.na(x))

# jarl-ignore-start any_is_na: this is a demo
...
any(is.na(x))
...
# jarl-ignore-end any_is_na

For more important changes to Jarl’s output, you will need to configure Jarl.

2. How to configure

2. How to configure


With CLI arguments (temporary) 

# Use the rules "any_is_na" and "any_duplicated" only
jarl check . --select any_is_na,any_duplicated

# Use all default rules + testthat rules except "expect_named"
jarl check . --extend-select TESTTHAT --ignore expect_named

# Get a more concise version of rule violations
jarl check . --output-format concise

# All commands
jarl check --help

2. How to configure


With jarl.toml (permanent) 

jarl.toml
[lint]

# general options
extend-select = ["TESTTHAT"]
ignore = ["implicit_assignment", "numeric_leading_zero"]

# rule-specific options
[lint.assignment]
operator = "="

[lint.unreachable_code]
extend-stopping-functions = ["my_custom_stop"]

CLI arguments always override jarl.toml arguments.

3. How to contribute

3. How to contribute


Requirements to add rules:

  • need some knowledge of Rust

  • need basic knowledge of Abstract Syntax Trees (AST)

  • no plugin mechanism, changes must be integrated in Jarl directly.


Note

I’m not a Rust expert, other code improvements are also welcome!

3. How to contribute


General contributing guide:

https://jarl.etiennebacher.com/contributing


Tutorial on contributing a rule:

https://jarl.etiennebacher.com/contributing-rule-example

3. What to contribute?



Some testthat rules to implement


Improve the coverage of lintr rules


Discuss existing suggestions, or propose your own rule!

Conclusion

Conclusion



You can install and run Jarl in three seconds (literally), give it a try!


Full docs: jarl.etiennebacher.com