ItemsJS - search engine in javascript

Extremely fast faceted search engine in JavaScript - lightweight, flexible, and simple to use. Created to perform fast search on json dataset (up to 100K items).

Demo

See another demo examples

Use cases

Itemsjs is being used mostly for data classification of companies, products, publications, documents, jobs or plants

The solution has been implemented by people from Amazon, Hermes, Apple, Microsoft, James Cook University, Carnegie Mellon University and more. You can find a list of real implementations - here

Features

• Ultra-fast faceted search: Process and filter data with blazing speed.
• Simple full-text search: Intuitive and straightforward text searching.
• Relevance scoring: Rank search results based on relevance.
• Facet filtering and sorting: Filter and order results by various facets.
• Pagination
• Works on both backend and frontend
• Integration with custom full-text search engines

Getting Started

NPM

npm install itemsjs

Using CommonJS syntax

const itemsjs = require('itemsjs')(data, configuration);
const items = itemsjs.search();

Using ES Module syntax

import itemsjs from 'itemsjs';
const searchEngine = itemsjs(data, configuration);
const items = searchEngine.search();

Client side

To use as an UMD in the browser:

<!-- CDN -->
<!-- unpkg: use the latest release -->
<script src="https://unpkg.com/itemsjs@latest/dist/index.umd.js"></script>
<!-- unpkg: use a specific version -->
<script src="https://unpkg.com/itemsjs@2.1.24/dist/index.umd.js"></script>
<!-- jsdelivr: use a specific version -->
<script src="https://cdn.jsdelivr.net/npm/itemsjs@2.1.24/dist/index.umd.js"></script>

<script>
  itemsjs = itemsjs(data, configuration);
  itemsjs.search()
</script>

To use as an ES module in the browser:

<!-- Include as ES Module -->
<script type="module">
  import itemsjs from 'https://unpkg.com/itemsjs@2.1.24/dist/index.module.js';
  // Initialize and use itemsjs here
  const searchEngine = itemsjs(data, configuration);
  searchEngine.search();
</script>

Example usage

npm install itemsjs

# download json data
wget https://raw.githubusercontent.com/itemsapi/itemsapi-example-data/master/items/imdb.json -O data.json

Next, create a search.js file with the following content:

const data = require('./data.json');

const itemsjs = require('itemsjs')(data, {
  sortings: {
    name_asc: {
      field: 'name',
      order: 'asc'
    }
  },
  aggregations: {
    tags: {
      title: 'Tags',
      size: 10,
      conjunction: false
    },
    actors: {
      title: 'Actors',
      size: 10
    },
    genres: {
      title: 'Genres',
      size: 10
    }
  },
  searchableFields: ['name', 'tags']
});

/**
 * get filtered list of movies 
 */
const movies = itemsjs.search({
  per_page: 1,
  sort: 'name_asc',
  // full text search
  // query: 'forrest gump',
  filters: {
    tags: ['1980s']
  }
})
console.log(JSON.stringify(movies, null, 2));

/**
 * get list of top tags 
 */
const top_tags = itemsjs.aggregation({
  name: 'tags',
  per_page: 10
})
console.log(JSON.stringify(top_tags, null, 2));

Run your script with Node.js:

node search.js

Integrations

If native full text search is not enough then you can integrate with external full text search.

How it works:

• each item of your data needs to have id field. It can be also custom field but it needs to be defined.
• native_search_enabled option in configuration should be disabled
• index data once in your search and itemsjs
• make search in your custom search and provide ids data into itemsjs
• done!

Examples:

• Integration with MiniSearch
• Integration with Lunr2.x

API

const itemsjs = ItemsJS(data, [configuration])

data

The first data argument is an array of objects.

configuration

Responsible for defining global configuration. Look for full example here - configuration

• aggregations filters configuration i.e. for tags, actors, colors, etc. Responsible for generating facets.

  Each filter can have it's own configuration. You can access those as buckets on the search() response.

  • title Human readable filter name
  • size Number of values provided for this filter (Default: 10)
  • sort Values sorted by count (Default) or key for the value name. This can be also an array of keys which define the sorting priority
  • order asc | desc. This can be also an array of orders (if sort is also array)
  • show_facet_stats true | false (Default) to retrieve the min, max, avg, sum rating values from the whole filtered dataset
  • conjunction true (Default) stands for an AND query (results have to fit all selected facet-values), false for an OR query (results have to fit one of the selected facet-values)
  • chosen_filters_on_top true (Default) Filters that have been selected will appear above those not selected, false for filters displaying in the order set out by sort and order regardless of selected status or not
  • hide_zero_doc_count true | false (Default) Hide filters that have 0 results returned

• sortings you can configure different sortings like tags_asc, tags_desc with options and later use it with one key.

• searchableFields an array of searchable fields.

• native_search_enabled if native full text search is enabled (true | false. It's enabled by default)

• isExactSearch set to true if you want to always show exact search matches. See lunr stemmer and lunr stopWordFilter.

• removeStopWordFilter set to true if you want to remove the stopWordFilter. See https://github.com/itemsapi/itemsjs/issues/46.

• fulltextSnapshot / facetsSnapshot optional prebuilt snapshots (from serializeAll or serializeFulltext/serializeFacets) to skip rebuilding indexes on cold start.

itemsjs.search(options)

options

• per_page amount of items per page.

• page page number - used for pagination.

• query used for full text search.

• sort used for sorting. one of sortings key

• filters filtering items based on specific aggregations i.e. {tags: ['drama' , 'historical']}

• filter function responsible for items filtering. The way of working is similar to js native filter function. See example

• filters_query boolean filtering i.e. (tags:novel OR tags:80s) AND category:Western

• is_all_filtered_items set to true if you want to return the whole filtered dataset.

• ids array of item identifiers to limit the results to. Useful when combining with external full-text search engines (e.g. MiniSearch).

Optional runtime facets (DX helper)

Instead of static filters you can pass facets with selections and runtime options (per-facet AND/OR, bucket size/sort):

const result = itemsjs.search({
  query: 'drama',
  facets: {
    tags: {
      selected: ['1980s', 'historical'],
      options: {
        conjunction: 'OR',      // AND/OR for this facet only (also accepts boolean true/false)
        size: 30,               // how many buckets to return
        sortBy: 'count',        // 'count' | 'key'
        sortDir: 'desc',        // 'asc' | 'desc'
        hideZero: true,         // hide buckets with doc_count = 0
        chosenOnTop: true,      // selected buckets first
      },
    },
  },
});
// response contains data.aggregations and an alias data.facets

facets is an alias/helper: under the hood it builds filters_query per facet (AND/OR) and applies bucket options. If you also pass legacy params, priority is: filters_query > facets > filters.

Ideal for React/Vue/Next UIs that need runtime toggles (AND/OR, “show more”, bucket sorting) without recreating the engine.

itemsjs.aggregation(options)

It returns full list of filters for specific aggregation

options

• name aggregation name
• per_page filters per page
• page page number
• query used for quering filters. It's not full text search
• conjunction true (Default) stands for an AND query, false for an OR query

itemsjs.similar(id, options)

It returns similar items to item for given id

options

• field field name for computing similarity (i.e. tags, actors, colors)
• minimum what is the minimum intersection between field of based item and similar item to show them in the result
• per_page filters per page
• page page number

itemsjs.reindex(data)

It's used in case you need to reindex the whole data

data

An array of objects.

Snapshots (optional)

Fast cold starts without reindexing. Snapshots are plain JSON, so you can store them wherever you like (localStorage, IndexedDB, file, CDN).

Generating a snapshot

const engine = itemsjs(data, config);
const snapshot = engine.serializeAll(); // { version, fulltext, facets }
// persist snapshot (e.g., localStorage / IndexedDB / file)

Using a snapshot

const snapshot = loadSnapshot(); // e.g., JSON.parse(...)
const engine = itemsjs(data, {
  ...config,
  fulltextSnapshot: snapshot.fulltext,
  facetsSnapshot: snapshot.facets,
});

APIs:

• itemsjs.serializeFulltext() → { index, store }
• itemsjs.serializeFacets() → { bitsData, ids, idsMap }
• itemsjs.serializeAll() → { version: 'itemsjs-snapshot-v1', fulltext, facets }

Snapshots are optional; if you don’t provide them, itemsjs rebuilds indexes as before.

Benchmarks

• See docs/benchmarks.md for snapshot/search benchmarks and the optional browser smoke test.