53 releases (13 breaking)

0.24.3 Oct 9, 2024
0.23.2 Oct 1, 2024
0.22.6 May 5, 2024
0.22.2 Mar 17, 2024
0.3.1 Jul 20, 2018

#1041 in Programming languages

Download history 51405/week @ 2024-07-18 52354/week @ 2024-07-25 50720/week @ 2024-08-01 52876/week @ 2024-08-08 50654/week @ 2024-08-15 52602/week @ 2024-08-22 48307/week @ 2024-08-29 57922/week @ 2024-09-05 59066/week @ 2024-09-12 55957/week @ 2024-09-19 59268/week @ 2024-09-26 63028/week @ 2024-10-03 61971/week @ 2024-10-10 65398/week @ 2024-10-17 54641/week @ 2024-10-24 71449/week @ 2024-10-31

266,792 downloads per month
Used in 562 crates (440 directly)

MIT license

620KB
15K SLoC

C 12K SLoC // 0.1% comments Rust 3.5K SLoC // 0.0% comments

Rust Tree-sitter

crates.io badge

Rust bindings to the Tree-sitter parsing library.

Basic Usage

First, create a parser:

use tree_sitter::{InputEdit, Language, Parser, Point};

let mut parser = Parser::new();

Add the cc crate to your Cargo.toml under [build-dependencies]:

[build-dependencies]
cc="*"

Then, add a language as a dependency:

[dependencies]
tree-sitter = "0.22"
tree-sitter-rust = "0.21"

To then use a language, you assign them to the parser.

parser.set_language(&tree_sitter_rust::language()).expect("Error loading Rust grammar");

Now you can parse source code:

let source_code = "fn test() {}";
let mut tree = parser.parse(source_code, None).unwrap();
let root_node = tree.root_node();

assert_eq!(root_node.kind(), "source_file");
assert_eq!(root_node.start_position().column, 0);
assert_eq!(root_node.end_position().column, 12);

Editing

Once you have a syntax tree, you can update it when your source code changes. Passing in the previous edited tree makes parse run much more quickly:

let new_source_code = "fn test(a: u32) {}";

tree.edit(&InputEdit {
  start_byte: 8,
  old_end_byte: 8,
  new_end_byte: 14,
  start_position: Point::new(0, 8),
  old_end_position: Point::new(0, 8),
  new_end_position: Point::new(0, 14),
});

let new_tree = parser.parse(new_source_code, Some(&tree));

Text Input

The source code to parse can be provided either as a string, a slice, a vector, or as a function that returns a slice. The text can be encoded as either UTF8 or UTF16:

// Store some source code in an array of lines.
let lines = &[
    "pub fn foo() {",
    "  1",
    "}",
];

// Parse the source code using a custom callback. The callback is called
// with both a byte offset and a row/column offset.
let tree = parser.parse_with(&mut |_byte: usize, position: Point| -> &[u8] {
    let row = position.row as usize;
    let column = position.column as usize;
    if row < lines.len() {
        if column < lines[row].as_bytes().len() {
            &lines[row].as_bytes()[column..]
        } else {
            b"\n"
        }
    } else {
        &[]
    }
}, None).unwrap();

assert_eq!(
  tree.root_node().to_sexp(),
  "(source_file (function_item (visibility_modifier) (identifier) (parameters) (block (number_literal))))"
);

Features

  • std - This feature is enabled by default and allows tree-sitter to use the standard library.
    • Error types implement the std::error:Error trait.
    • regex performance optimizations are enabled.
    • The DOT graph methods are enabled.
  • wasm - This feature is enabled for Wasm targets. tree-sitter to be built for Wasm targets using the wasmtime-c-api crate.

Dependencies

~2–15MB
~205K SLoC