shiki_flutter_dart_engine

A pure-Dart Oniguruma engine backend for shiki_flutter, wired through the oniguruma_dart port: a 1:1 Dart implementation of the Oniguruma regex engine. No FFI, no native build: it runs on every platform, web included.

It exists because shiki_flutter's engine is pluggable: ShikiHighlighterEngine is a seam the TextMate tokenizer runs on, and this package provides an implementation backed by a faithful Oniguruma port.

Installation

dependencies:
  shiki_flutter_dart_engine: ^1.0.0

Usage

It's already the default on native/VM: shiki_flutter depends on this package and uses ShikiHighlighterDartEngine as its default engine on native and the Dart VM (faithful semantics, and it edges out the bundled engine there). On web it falls back to the built-in ShikiHighlighterEmbeddedEngine, whose RegExp fast path is several times faster than the port under dart2js. So for that recommended per-platform setup you don't have to write anything: importing shiki_flutter is enough.

To force the port on every platform (web included), set it explicitly:

import 'package:shiki_flutter/shiki_flutter.dart';
import 'package:shiki_flutter_dart_engine/shiki_flutter_dart_engine.dart';

void main() {
  ShikiHighlighter.config = const ShikiHighlighterConfig(
    ioEngine: ShikiHighlighterDartEngine(),
    webEngine: ShikiHighlighterDartEngine(),
  );
  runApp(const MyApp());
}

Or per-highlighter: ShikiHighlighter(engine: const ShikiHighlighterDartEngine()).

Which engine should I use?

Reach for this engine when you want Oniguruma semantics without the native build/FFI constraints of shiki_flutter_native_engine, including on web, where FFI isn't available. Unlike that engine, the port drives Oniguruma in UTF-8, so 2-digit \xHH escapes (e.g. CSS's [^\x00-\x7F]) compile correctly; offsets are still reported in UTF-16 code units, matching Dart String indices.

Correctness

test/parity_test.dart tokenizes the core package's golden fixtures (proven byte-identical to real Shiki) with both this engine and the built-in pure-Dart engine and asserts the token streams are identical, across all sampled languages and themes.

Performance note

The scanner compiles each grammar pattern to its own OnigRegex and searches them one by one, picking the left-most match. The input line is encoded once per pattern (cached by the port on the line's String identity) rather than once per scanner. A future optimization can switch to the port's OnigRegSet to encode once and search all patterns together. On web the built-in ShikiHighlighterEmbeddedEngine (the default there) is faster than the port, so there's usually no reason to force this engine on web.