inditrans 0.17.4

A functional transliterator library for the most common Indic/Brahmic scripts.

inditrans

A transliteration library for Indic/Brahmic and Roman scripts for Flutter and Dart. Uses FFI on native platforms and WASM on Web for near-native performance, with no server dependency.

• Available on all platforms (Android, iOS, Linux, macOS, Windows, Web).
• ITRANS toggle transliteration — wrap any region in ##…## to pass it through unchanged.
• XML/HTML tag pass-through — tags are never transliterated.
• Special handling for Tamil, Malayalam, Gurmukhi, and other scripts tuned for readability.
• Vedic accent support for Vedic scriptures.

Supported scripts

Optional Urdu pronunciation

transliterateUrduWithLexicon(text, to, bytes, options) accepts a caller-supplied IUL1 or IUL2 pack as Uint8List to override the bundled experimental dictionary. Ordinary Urdu conversion uses the bundled dictionary without downloading data. A single reading is selected without contextual disambiguation. Unknown, explicitly marked and mixed-script words use the ordinary Urdu reader. DisableInference bypasses the lexicon. See lexicon compilation and runtime behavior for the pack format and current limitations.

Indic/Brahmic: Assamese, Bengali, Burmese, Devanagari, Grantha, Gujarati, Gurmukhi, Kannada, Khmer, Malayalam, Oriya, Sinhala, Tamil, Telugu, Thai, Tibetan

Roman schemes: IAST · IPA · ISO 15919 · ITRANS · TITUS · Indlish (my own take on a readable Roman representation for Indian scripts)

ITRANS notation notes

inditrans implements uniform phonemic vowel mapping adhering to the ITRANS specification:

• Short vowels: e and o encode short vowels (e.g. Kannada ಎ / ೆ, Telugu ఎ / ె, Tamil எ / ெ, Malayalam എ / െ, Devanagari ऎ / ॆ). ^e and ^o remain supported as aliases.
• Long vowels: E and O encode long vowels (e.g. Kannada ಏ / ೇ, Telugu ఏ / ే, Tamil ஏ / ே, Malayalam ഏ / േ, Devanagari ए / े).
• Examples: te $\to$ तॆ / ತೆ, tE $\to$ ते / ತೇ; tere bina $\to$ ತೆರೆ ಬಿನ (Kannada), tErE bina $\to$ ತೇರೇ ಬಿನ (Kannada) / तेरे बिन (Devanagari).

Requirements

• Flutter ≥ 3.38.10 / Dart SDK ^3.10.8
• No additional runtime dependencies for transliteration

Installation

flutter pub add inditrans

Quick start

import 'package:inditrans/inditrans.dart' as inditrans;

Future<void> main() async {
  // init() must complete before transliterate() is called
  await inditrans.init();

  final result = inditrans.transliterate(
    "श्री॒ गु॒रु॒भ्यो नमः॒ । ह॒रिः॒ ओ३म् ॥",
    inditrans.Script.devanagari,
    inditrans.Script.tamil,
    inditrans.Options.IgnoreVedicAccents + inditrans.Options.TamilSuperscripted,
  );
  print(result);
}

Configuration

Pass an Option value (or combine with +) as the fourth argument to transliterate():

Option	Description
Option.TamilTraditional	Traditional Tamil consonants only
Option.TamilSuperscripted	Superscripted Tamil consonants
Option.DevanagariTraditional	Traditional Devanagari orthography (maps Dravidian short e/o and consonants)
Option.ASCIINumerals	Force ASCII numerals in output
Option.IgnoreVedicAccents	Strip Vedic accent markers
Option.RetainSpecialMarkers	Keep non-standard character markers
Option.NoXMLTagHandling	Treat XML/HTML tags as plain text
Option.DisableInference	Disable anuswara assimilation and contextual Tamil pronunciation/letter inference

For Tamil source text, Option.TamilSuperscripted also enables DisableInference automatically; using it only for Tamil output does not. With DisableInference, a target's mapped anuswara is emitted instead of a contextually inferred nasal (which may differ in spelling).

The skipStart and skipEnd parameters (default "##") mark regions that are not transliterated and are passed through unchanged.

Development

flutter pub get          # install Dart dependencies
dart analyze             # lint and static analysis
flutter test             # run tests (native library must be built)
python3 ../scripts/generate_bindings.py  # regenerate FFI bindings after native API changes

Checkout builds use the shared engine in ../native/src/. Before publishing, run python3 scripts/prepare_flutter_package.py from the repository root to bundle it in flutter/src/ and copy the changelog. Installed pub.dev packages build from those bundled sources. Verify that layout with python3 tests/integration/flutter_package_smoke.py from the repository root.

Documentation

• Changelog
• Project on GitHub
• Node.js / JavaScript package

License

MIT

Forward accuracy and Urdu

Script.urdu (alias Script.nastaliq) supports Urdu/Perso-Arabic text. Use a Nastaliq-capable font and RTL text direction in your application. The returned string is ordinary Unicode in logical order. Explicit vowels, aspirates and shadda are supported; ordinary unvocalized Urdu needs lexical information to recover omitted vowels, so bare consonants receive no guessed vowel.

Targets use their native letters or documented phonetic approximations; source distinctions may merge. ISO input now accepts the additional spreadsheet notations, including ê, ŭ, ĉ, ž, w and the Urdu loan-letter spellings. Artificial extended-vowel separators/accents have been replaced by native approximations. Real supported accents remain separate marks. There is no round-trip guarantee.

The script profiles implement transliteration, not complete language-specific pronunciation. See the repository's mapping review for exact notation behavior and limitations.

Reusable Urdu pronunciation lexicon (experimental)

Call UrduPronunciationLexicon.load(iulBytes) after init(). It returns null for invalid bytes. Call lexicon.transliterate(text, Script.iast) for repeated conversion, then lexicon.dispose() when finished. The handle owns decoded data and no longer needs the source bytes. The existing one-shot transliterateUrduWithLexicon() remains available.