@squawk/search

Domain-agnostic fuzzy string matching and ranked search scoring. A small, dependency-free engine that scores how well a query matches a candidate string (exact, prefix, substring, subsequence, and bounded typo tiers), returns a normalised [0, 1] score plus the matched character ranges for UI highlighting, and ranks a list of items across one or more searchable fields.

Documentation

Part of the @squawk aviation library suite. See all packages on npm.

This package holds no aviation domain knowledge. It is the shared scoring substrate used by the @squawk/* resolvers (airports, navaids, fixes, airways, airspace) to power their search() methods, and can be used standalone for any fuzzy-matching need.

Installation

npm install @squawk/search
Copy

Usage

Score a single candidate

fuzzyScore compares a query against one candidate string and returns a normalised score in [0, 1] (1 = exact, 0 = no match) along with the matched character ranges:

import { fuzzyScore } from '@squawk/search';

fuzzyScore('jfk', 'JFK'); // { score: 1, ranges: [{ start: 0, end: 3 }] }
fuzzyScore('ken', 'Kennedy'); // strong prefix match, ranges cover "Ken"
fuzzyScore('kndy', 'Kennedy'); // subsequence match, lower score
fuzzyScore('jfx', 'JFK'); // bounded typo match, lower score still
fuzzyScore('xyz', 'JFK'); // { score: 0, ranges: [] }
Copy

Rank a list of items

fuzzySearch ranks a list of arbitrary items against a query. You describe how to extract the searchable fields from each item; the engine scores every field, keeps the best per item, filters, sorts by descending score, and slices to limit:

import { fuzzySearch } from '@squawk/search';

interface Station {
  id: string;
  name: string;
}

const stations: Station[] = [
  { id: 'JFK', name: 'John F Kennedy Intl' },
  { id: 'EWR', name: 'Newark Liberty Intl' },
  { id: 'LGA', name: 'LaGuardia' },
];

const matches = fuzzySearch(stations, 'kennedy', {
  keys: (s) => [
    { name: 'id', text: s.id },
    { name: 'name', text: s.name },
  ],
  limit: 10,
});

// matches[0] => { item: { id: 'JFK', ... }, score, field: 'name', ranges: [...] }
Copy

Key Features

• Tiered scoring: Exact, prefix, word-boundary prefix, substring, subsequence, and bounded Damerau-Levenshtein typo tiers, collapsed into a single normalised [0, 1] score so results are comparable across fields and data sets.
• Match ranges: Every score carries the matched character ranges, ready for highlight rendering in a UI.
• Multi-field ranking: Score an item across several fields and keep the best-matching one, with the matched field name reported on each result.
• Caller-supplied filtering: Pass a predicate to exclude items before scoring (e.g. respecting visibility or type filters) without the engine knowing anything about the domain.
• Zero dependencies: Pure string logic with no runtime dependencies.