Skip to contents

astgrepr

astgrepr provides R bindings to the ast-grep Rust crate. ast-grep is a tool to parse the abstract syntax tree (AST) of some code and to perform search and rewrite of code. This is extremely useful to build linters, stylers, and perform a lot of code analysis.

See the example below and the “Getting started” vignette for a gentle introduction to astgrepr.

Since astgrepr can be used as a low-level foundation for other tools (such as linters), the number of R dependencies is kept low:

> pak::local_deps_tree()
✔ Loading metadata database ... done
local::. 0.0.1 [new][bld]                                                  
├─checkmate 2.3.1 [new][dl] (746.54 kB)
│ └─backports 1.5.0 [new]
├─rrapply 1.2.7 [new]
└─yaml 2.3.8 [new][dl] (119.08 kB)

Key:  [new] new | [dl] download | [bld] build

Installation 

install.packages('astgrepr', repos = c('https://etiennebacher.r-universe.dev', 'https://cloud.r-project.org'))

Demo 

library(astgrepr)

src <- "library(tidyverse)
x <- rnorm(100, mean = 2)
any(duplicated(y))
plot(x)
any(duplicated(x))
any(is.na(variable))"

root <- src |>
  tree_new() |>
  tree_root()

# get everything inside rnorm()
root |>
  node_find(ast_rule(pattern = "rnorm($$$A)")) |>
  node_get_multiple_matches("A") |>
  node_text_all()
#> $rule_1
#> $rule_1[[1]]
#> [1] "100"
#> 
#> $rule_1[[2]]
#> [1] ","
#> 
#> $rule_1[[3]]
#> [1] "mean = 2"

# find occurrences of any(duplicated())
root |>
  node_find_all(ast_rule(pattern = "any(duplicated($A))")) |>
  node_text_all()
#> $rule_1
#> $rule_1$node_1
#> [1] "any(duplicated(y))"
#> 
#> $rule_1$node_2
#> [1] "any(duplicated(x))"

# find some nodes and replace them with something else
nodes_to_replace <- root |>
  node_find_all(
    ast_rule(id = "any_na", pattern = "any(is.na($VAR))"),
    ast_rule(id = "any_dup", pattern = "any(duplicated($VAR))")
  )

fixes <- nodes_to_replace |>
  node_replace_all(
    any_na = "anyNA(~~VAR~~)",
    any_dup = "anyDuplicated(~~VAR~~) > 0"
  )

# original code
cat(src)
#> library(tidyverse)
#> x <- rnorm(100, mean = 2)
#> any(duplicated(y))
#> plot(x)
#> any(duplicated(x))
#> any(is.na(variable))

# new code
tree_rewrite(root, fixes)
#> library(tidyverse)
#> x <- rnorm(100, mean = 2)
#> anyDuplicated(y) > 0
#> plot(x)
#> anyDuplicated(x) > 0
#> anyNA(variable)

There is some recent work linking tree-sitter and R. Those are not competing with astgrepr but are rather a complement to it:

  • r-lib/tree-sitter-r: provide the R grammar to be used with tools built on tree-sitter. astgrepr relies on this grammar under the hood.
  • DavisVaughan/r-tree-sitter: a companion of r-lib/tree-sitter-r. This gives a way to get the tree-sitter representation of some code directly in R. This is useful to learn how tree-sitter represents the R grammar, which is required if you want advanced use of astgrepr. However, it doesn’t provide a way to easily select specific nodes (e.g. based on patterns).