flir

(pronounced “fleer” as in “beer”)

flir is an R package to detect and rewrite code patterns. It was originally created to be an R linter and an alternative to lintr. However, it may be better to view it as a tool to refactor any type of code by detecting and rewriting custom patterns (see Adding new rules). flir comes with a list of built-in rules and therefore can still be used as a linter (see the “Usage” section below), but I now concentrate my efforts on a new R linter entirely written in Rust: Jarl. Therefore, I will not add new rules in flir.

flir is powered by astgrepr, which is itself built on the Rust crate ast-grep.

Installation

To get the CRAN version (stable):

install.packages("flir")

To get the development version (unstable):

# install.packages("remotes")
remotes::install_github("etiennebacher/flir")

Usage

Optional setup:

• setup_flir(): creates the folder flir and populates it with built-in rules as well as a cache file. You can modify those rules or add new ones if you want more control.

You can use flir as-is, without any setup. However, running setup_flir() enables the use of caching, meaning that the subsequent runs will be faster. It is also gives you a place where you can store custom rules for your project/package.

The everyday usage consists of two functions:

• lint() looks for rule violations in R files;
• fix() looks for rule violations in R files and automatically applies their replacement (if any).

One can also experiment with flir::lint_text() and flir::fix_text():

flir::lint_text(
  "
any(is.na(x))
any(duplicated(y))
"
)
#> Original code: any(is.na(x)) 
#> Suggestion: anyNA(x) is better than any(is.na(x)). 
#> Rule ID: any_na-1 
#> 
#> Original code: any(duplicated(y)) 
#> Suggestion: anyDuplicated(x, ...) > 0 is better than any(duplicated(x), ...). 
#> Rule ID: any_duplicated-1
flir::fix_text(
  "
any(is.na(x))
any(duplicated(y))
"
)
#> Old code:
#> any(is.na(x))
#> any(duplicated(y))
#> 
#> New code:
#> anyNA(x)
#> anyDuplicated(y) > 0

See the vignette Automatic fixes to see how to be more confident about changes introduced by flir.

Real-life examples

I tested flir on several packages while developing it. I proposed some pull requests for those packages. Here are a few:

• ggplot2: #6050 and #6051
• marginaleffects: #1171 and #1177
• targets: #1325
• tinytable: #325
• usethis: #2048

Except for some manual tweaks when the replacement was wrong (I was testing flir after all), all changes were generated by flir::fix_package() or flir::fix_dir(<dirname>).

Comparison with existing tools

The most used tool for lints detection in R is lintr. However, lintr’s performance is not optimal when it is applied on medium to large packages. Also, lintr cannot perform automatic replacement of lints.

styler is a package to clean code by fixing indentation and other things, but doesn’t perform code replacement based on lints.

flir is quite fast This is a small benchmark on 3.5k lines of code with a few linters:

file <- system.file("bench/test.R", package = "flir")

bench::mark(
  lintr = lintr::lint(
    file,
    linters = list(
      lintr::any_duplicated_linter(),
      lintr::any_is_na_linter(),
      lintr::matrix_apply_linter(),
      lintr::function_return_linter(),
      lintr::lengths_linter(),
      lintr::T_and_F_symbol_linter(),
      lintr::undesirable_function_linter(),
      lintr::expect_length_linter()
    )
  ),
  flir = flir::lint(
    file,
    linters = list(
      flir::any_duplicated_linter(),
      flir::any_is_na_linter(),
      flir::matrix_apply_linter(),
      flir::function_return_linter(),
      flir::lengths_linter(),
      flir::T_and_F_symbol_linter(),
      flir::undesirable_function_linter(),
      flir::expect_length_linter()
    ),
    verbose = FALSE,
    open = FALSE
  ),
  check = FALSE
)
#> Warning: Some expressions had a GC in every iteration; so filtering is disabled.
#> # A tibble: 2 × 6
#>   expression      min   median `itr/sec` mem_alloc `gc/sec`
#>   <bch:expr> <bch:tm> <bch:tm>     <dbl> <bch:byt>    <dbl>
#> 1 lintr         3.44s    3.44s     0.291   313.5MB    12.2 
#> 2 flir       153.77ms 172.76ms     5.89      1.8MB     1.96

Why the name “flir”?

flir was originally named flint but I had to rename it to avoid conflicts with a package named flint on CRAN.

flir stands for “Fix Lints In R”.

Contributing

Did you find some bugs or some errors in the documentation? Do you want flir to support more rules?

Take a look at the contributing guide for instructions on bug report and pull requests.

Acknowledgements

The website theme was heavily inspired by Matthew Kay’s ggblend package: https://mjskay.github.io/ggblend/.