highlight — syntax highlighting spans

A small, self-contained highlighter: given source text and an explicit language name, it returns the byte ranges of tokens worth coloring.

pub fn highlight(source: &str, lang: &str) -> Option<Vec<HighlightSpan>>

Unlike the rest of the module, highlight takes a language name (not a path) and does not use the LRU cache — it is a one-shot parse for rendering snippets (code blocks in markdown, terminal output).

Language names

lang is matched against a fixed set of names. Note the JS family collapse: javascript, js, jsx, mjs, cjs, json, typescript, ts, tsx, php, lua, and dart all use the JavaScript grammar (with the JS query). Matching is exact — a leading/trailing space (" rust", "rust ") returns None (this mirrors how a markdown renderer passes code-fence languages and rejects stray whitespace).

Unknown names return None.

Categories and queries

The output is a sorted list of HighlightSpan { start, end, category } (byte offsets into source):

pub enum HighlightCategory { Keyword, String, Comment, Type, Function, Number, Builtin }

Each language has a hand-written query:

• Rust, Python, JavaScript get full queries: keywords (literals and structural words), strings, comments, types, numbers, function definitions and call sites, and builtins (true/false/null/ this/self/…).
• The remaining languages (C#, Go, Java, Scala, Haskell, Swift, Zig, Kotlin, C/C++, Objective-C, …) share a minimal fallback query covering just strings, comments, types, and numbers.

Spans are sorted by start; overlapping captures are possible (e.g. a keyword inside a comment) — consumers decide how to resolve them.

Example

let spans = tree_sitter::highlight::highlight("fn main() { let x = 1; }", "rust");
for span in spans.unwrap() {
    let text = &src[span.start..span.end];
    println!("{:?}: {text:?}", span.category);
}
// Keyword: "fn", Keyword: "let", Number: "1", ...