esjs

Elasticsearch-ish inverse indexed search engine for Node.

Quickstart

Install

yarn add https://github.com/10eTechnology/esjs

Use

import ESjs from 'esjs';

const fields = {
  name:        { boost: 2 },
  description: { boost: 1 },
  job:         null,
  status:      { analyzer: 'keyword' },
};

const idx = new ESjs({ fields });

idx.addDoc({
  id: 1,
  name: 'Larry Jones',
  description: 'A nice guy',
  job: 'plumber',
  status: 'AVAILABLE',
});

idx.addDoc({
  id: 2,
  name: 'Moe Jones',
  description: "Larry's brother, a decent guy",
  job: 'architect',
  status: 'AVAILABLE',
});

idx.addDoc({
  id: 2,
  name: 'Curly Jones',
  description: "Moe's brother, a funny guy",
  job: 'mathematician',
  status: 'UNAVAILABLE',
});

idx.search('Larry');
idx.search({ must: { match: { _all: 'Larry' } } }); // equivalent
// [{ id: 1, score: 0.1 }, { id: 2, score: 0.05 }] (scores are not real)
idx.search({ must: { match: { name: 'Larry' } } });
// [{ id: 1, score: 0.9 }]
idx.search({
  must: {
    match: {
      name:        { query: 'Larry', boost: 5 },
      description: { query: 'guy', boost: 1 },
    },
  },
});
// [{ id: 1, score: 0.1 }, { id: 2, score: 0.05 }, { id: 3, score: 0.05 }]
idx.search({
  must: {
    match: { description: 'brother' },
    term:  { job: 'plumber' },
  },
});
// [{ id: 1, score: 0.1 }, { id: 2, score: 0.01 }, { id: 3, score: 0.01 }]
idx.search({
  must: {
    match: {
      name: 'Jones',
      description: 'brother',
    },
  },
  filter: [{
    term: { job: 'plumber' }
  }],
});
// []
idx.search({
  must: {
    match: {
      name: 'Jones',
    },
  },
  filter: [{
    term: { status: 'AVAILABLE' }, 
  }],
});
// [{ id: 1, score: 0.1 }, { id: 2, score: 0.1 }]
idx.search('AVAILABLE');
// []

const json = idx.serialize();
// JSON string suitable for persisting somewhere
serializedIndex = new idx(null, json);

serializedIndex.search('Larry');

You can save documents in the index and they will be returned in the results data:

const idx = new ESjs({ fields, storeDocs: true });

// add some docs, then

idx.search('Larry');
// [{ id: 1, score: 0.9, doc: {...} }]

Partial string matches

You can configure ESjs to search partial strings (from the beginning of each) word, at the cost of a greatly increased index size.

Partial matches are penalized during the scoring phase, causing them to appear below matches on full words.

const idx = new ESjs({ fields, allowPartial: true });

// add some docs, then

idx.search('lar');
// [{ id: 1, score: 0.3 }]

Turning off stopwords

You can turn off stopwords by setting stopwords: false in your configuration.

This is a temporary implementation. Configuring the tokenizer pipeline will improve with time*

Field analyzer

You may specify a custom analyzer per field with the analyzer config option.

const fields = {
  status: { analyzer: 'keyword' },
};

Currently, the only analyzer supported is keyword or standard. This option defaults to standard.

standard

Default behavior. Field will be searchable through text search or term filters.

keyword

The keyword analyzer will not index specified field for text searching. The only way to surface results containing a keyword field is through the an exact match on a filter term query configuration.