urdu_tools #

A comprehensive, zero-dependency, and strongly-typed Dart library for advanced Urdu text processing. This package is part of the urdu-tools ecosystem (supporting TypeScript, C#, and Dart).

It includes modules for:

• Text Normalization (Diacritics, Kashida, Hamza, Zero-Width joiners)
• Text Analysis (RTL detection, Character classification)
• Compound Word Detection (Prefixes, Suffixes, Izafat, Lexicon-based matching)
• Legacy Encoding translation (InPage v1/v3 decoding, Windows-1256 to Unicode)
• Tokenization (Sentences, N-grams, Unique tokens)
• Number System Parsing (Words to numbers, Numbers to words)
• Sorting and Collation (Proper 39-letter Urdu collation)
• Transliteration (Urdu to Roman, Roman to Urdu)

Installation #

Add this to your package's pubspec.yaml file:

dependencies:
  urdu_tools: ^1.4.2

Quick Start #

import 'package:urdu_tools/urdu_tools.dart';

void main() {
  // 1. Text Normalization
  final uglyText = 'ب\u064Eس\u0651م اﷲ'; // Fatha, Shadda, Allah ligature
  final clean = normalize(uglyText);
  print(clean); // 'بسم اللہ' (Standardized)

  // 2. Number Parsing
  print(wordsToNumber('ایک سو پچیس')); // 125
  print(numberToWords(125)); // 'ایک سو پچیس'

  // 3. Compound Word Detection
  final spans = detectCompounds('یہ کتب خانہ اور طالبِ علم ہے۔');
  for (var span in spans) {
    print('${span.text} [${span.type.name}]'); 
    // کتب خانہ [affix]
    // طالبِ علم [izafat]
  }

  // 4. Tokenization & NLP
  final tokens = tokenize('یہ ایک، جملہ ہے۔');
  print(tokens.map((t) => t.text).toList()); // ['یہ', 'ایک', 'جملہ', 'ہے']

  // 5. InPage Legacy Decoding
  // Decode legacy InPage byte arrays (.inp files) directly to Unicode!
  final legacyBytes = [0x04, 0x83, 0x20]; // [InPage v1 Alef, Peh, Space]
  final decoded = decodeInpage(legacyBytes);
  print(decoded);
}

Features #

• Zero dependencies: Extremely lightweight. Uses only core dart:typed_data for memory efficiency.
• Strictly Typed: Extensive enums and data structures for robust, crash-free performance in Flutter apps.
• Cross-platform parity: Functionally identical to the TypeScript (urdu-js) and C# (urdu-dotnet) versions.
• Tested: Covered by 120+ unit tests modeling real-world Urdu edge cases.

Contributing #

We welcome contributions! Please note that the compound word lexicon dataset is synchronized across multiple languages. Please see the main CONTRIBUTING.md guide at the repository root for rules on submitting changes to the dictionaries.