ja-k-e / fuzzy-search

A JavaScript plugin to find exact and fuzzy matches in a library of strings.

Fuzzy Search

A JavaScript plugin to find exact and fuzzy matches in a library of strings.

Using Generator Babel Boilerplate.

Example

codepen.io/jakealbaugh/pen/wzzrmX

Overview

Fuzzy Search provides search results from a search term (term) and an array of strings or objects (library). FS.search('lo', ['hello', 'lingo']).

Exact / Fuzzy

So that you can prioritize types of matches, fuzzy search returns two arrays of matches, exact and fuzzy. exact matches contain the exact search term, ordered by proximity to the beginning of the string in which it was found. fuzzy matches contain the search term even if there are characters in between.

For example, if I search for 'jake' and have two strings to search, ['jakealbaugh', 'jackeagle'], there is an exact match 'jakealbaugh' and fuzzy match 'jackeagle'.

Basic Usage

In its most basic form, FS.search uses a search term and a simple array of strings.

Input

FS.search('lo', ['hello', 'lingo']);

Output

{
  success: true,
  count: 2,
  term: 'lo',
  exact: [
    {
      id: 1,
      string: 'hello',
      _matchType: 'exact',
      _substrings: [
        { match: false, str: 'hel' },
        { match: true,  str: 'lo' }
      ],
      _score: 0
    }
  ],
  fuzzy: [
    {
      id: 2,
      string: 'lingo',
      _matchType: 'fuzzy',
      _substrings: [
        { match: true,  str: 'l' },
        { match: false, str: 'ing' },
        { match: true,  str: 'o' }
      ],
      _score: 1
    }
  ],
  _regex: {
    exact: '(.+)?(lo)(.+)?$',
    fuzzy: '(.+)?(l)(.+)?(o)(.+)?$'
  }
}

Substrings

As you can see, a match has a handful of properties returned with it as well. The most important of these properties is the _substrings array. What good is a match if you have no way of displaying it? You can use this value to display the match in a UI by highlighting the substrings with a match value of true.

Match _score is relative to the match type, so the score of a fuzzy match has no relation to the score of an exact one.

Usage with Objects

Chances are your library requires some sort of identification or has other properties. FS will create basic ids for you if you pass in a plain array of strings, but you can use objects if you want to preserve your own data.

The base schema for a fuzzy object is an id and string value. You can leave out id if you don't care or want to handle identification yourself, but a string value is required and must be the term that FS searches. You can pass around any other attributes you want with the fuzzySearch object, but the result will overwrite reserved attributes (denoted by an _).

Input

var string_lib = [
  { id: 123, string: 'hello', YOUR_ATTR: false },
  { id: 456, string: 'lingo', YOUR_ATTR: false }
];
FS.search('lo', string_lib);

The output maintains the fluff and adds the rest of the report:

Output

{
  success: true,
  count: 2,
  term: 'lo',
  exact: [
    {
      id: 123,
      string: 'hello',
      YOUR_ATTR: false,
      _matchType: 'exact',
      _substrings: [
        { match: false, str: 'hel' },
        { match: true,  str: 'lo' }
      ],
      _score: 0
    }
  ],
  fuzzy: [
    {
      id: 456,
      string: 'lingo',
      YOUR_ATTR: false,
      _matchType: 'fuzzy',
      _substrings: [
        { match: true,  str: 'l' },
        { match: false, str: 'ing' },
        { match: true,  str: 'o' }
      ],
      _score: 1
    }
  ],
  _regex: {
    exact: '(.+)?(lo)(.+)?$',
    fuzzy: '(.+)?(l)(.+)?(o)(.+)?$'
  }
}

fuzzy-search's Issues

improved scoring

currently, scores aren't that impressive. exact match score only denotes that a match is at the beginning of a string or isn't. that could be improved.

case insensitivity

improved fuzzy regex

the regex generated for fuzzy favors distance between chars, could also benefit from a * instead of +?.