tree-sitter.zig ​

A native Zig implementation of the Tree-sitter parsing runtime.

Fast, incremental, dependency-free parsing for Zig 0.16.0 — no C code, no Rust, no @cImport. Just Zig and its standard library.

What is tree-sitter.zig? ​

Tree-sitter is a parsing strategy built for developer tools: instead of parsing a file once and throwing the result away, it keeps a concrete syntax tree alive, updates it cheaply on every keystroke, and lets tools ask structural questions about code.

tree-sitter.zig implements that runtime natively in Zig:

const treesitter = @import("treesitter");

var parser = treesitter.Parser.init(allocator);
defer parser.deinit();
try parser.setLanguage(my_language);

var tree = try parser.parseString("1 + 2 * 3");
defer tree.deinit();

const root = tree.rootNode(); // lightweight handle, no allocation

Features ​

Quick Start ​

var parser = treesitter.Parser.init(allocator);
defer parser.deinit();
try parser.setLanguage(my_language);

var tree = try parser.parseString("a + b");
defer tree.deinit();

var query = try parser.compileQuery("(identifier) @id");
defer query.deinit();

var qcursor = parser.queryCursor();
defer qcursor.deinit();
try qcursor.execute(my_language, query.patterns(), query.nodes(), query.captureNames(), &tree);
while (qcursor.nextMatch()) |m| {
    for (m.captures) |cap| std.debug.print("@{s}: {s}\n", .{ cap.name, cap.node.text() });
}

Read the Getting Started guide, browse runnable examples, or jump to the API reference.

Architecture ​

                    tree-sitter.zig
                           │
        ┌──────────────────┼──────────────────┐
        │                  │                  │
      Parser             Tree              Query
        │                  │                  │
   ┌────┼────┐       ┌─────┼─────┐       ┌────┼────┐
   │    │    │       │     │     │       │    │    │
 Lexer Stack Actions Node Cursor Edit  Pattern Capture Matcher
   │
 Input
   │
 Unicode
   │
 Language

See Architecture for the full picture.

Core Concepts ​

• Parser — turns source text into a syntax tree.
• Nodes — lightweight handles into the tree.
• Cursors — allocation-free tree traversal.
• Incremental parsing — reparse only what changed.
• Queries — structural search with captures.

Use Cases ​

From syntax highlighting to editor integration and static analysis — if your tool reads code structurally, start with the use cases.

Performance ​

On an ~80KB expression corpus (100,000 nodes, ReleaseFast): initial parse ~10 ms, worst-case incremental edit ~4–16 ms with ~40,000 reused subtrees, traversal ~1 ms, 20,000 query matches in ~4 ms. Details and reproduction steps: Performance.

Compatibility ​

Windows, Linux, and macOS across x86, x86_64, and ARM64 — validated with Zig 0.16.0. See the compatibility matrix.

GitHub & Community ​

• Repository: muhammad-fiaz/tree-sitter.zig
• Issues and feature requests welcome.
• If this project helps you, please star it.