string_similarity

A fast, dependency-free Dart package for measuring the similarity between two strings with the Sørensen–Dice coefficient.

Use it for fuzzy search, spelling suggestions, duplicate detection, and ranking the closest match from a list.

• Simple extension methods and static methods
• Scores from 0.0 (completely different) to 1.0 (identical)
• Best-match search with individual ratings and the winning index
• No runtime dependencies
• Works on every platform supported by Dart and Flutter

dart pub add string_similarity

Then import the package:

import 'package:string_similarity/string_similarity.dart';

final score = 'healed'.similarityTo('sealed');

print(score); // 0.8

The equivalent static method is:

final score = StringSimilarity.compareTwoStrings('healed', 'sealed');

final result = 'healed'.bestMatch([
  'edward',
  'sealed',
  'theatre',
]);

print(result.bestMatch.target); // sealed
print(result.bestMatch.rating); // 0.8
print(result.bestMatchIndex); // 1

Every candidate and its score remain available in result.ratings.

The equivalent static method is:

final result = StringSimilarity.findBestMatch(
  'healed',
  ['edward', 'sealed', 'theatre'],
);

The package compares adjacent character pairs (bigrams) using the Sørensen–Dice coefficient. Whitespace is removed before comparison. Letter case and diacritics are kept, so Cote, cote, and côte are different values.

Normalize both strings before comparing them:

final score = 'DART'.toLowerCase().similarityTo('dart'.toLowerCase());

print(score); // 1.0

BestMatch provides:

• ratings: every target and its score
• bestMatch: the highest-rated target
• bestMatchIndex: the winning target's original index

• Whitespace is ignored, but punctuation is not.
• Comparison is case-sensitive and diacritic-sensitive.
• The algorithm works on UTF-16 code units rather than grapheme clusters, so some emoji and combined Unicode characters may not behave like visible characters.
• Sørensen–Dice measures shared bigrams. It is not a drop-in replacement for edit-distance algorithms such as Levenshtein distance.

Licensed under the permissive MIT License.