Picosearch NPM

picosearch

Warning: As long as the version is in the 0.X.X range; version changes are most likely breaking!

Minimalistic, customizable module for creating basic full-text search indices and queries using the BM25F algorithm (used by Lucene, Elasticsearch etc.). The focus is on providing a simple and reusable implementation and configuration with no dependencies.

Stemmers and stopwords are not included and must be provided as config values.
JSON serializable indices
Highly customizable

Installation

yarn add picosearch

npm install picosearch

Quickstart

const { createIndex, indexDocument, searchIndex } = require('picosearch')
const porterStemmer = require('porter-stemmer')
const { eng } = require('stopword')

; (async () => {
  // define a (custom) tokenizer for splitting a sentence into tokens
  const tokenizer = (sentence) => sentence.split(' ').map(s => s.trim())

  // define a (custom) anaylzer for preprocessing individual tokens/words
  const REGEXP_PATTERN_PUNCT = new RegExp("['!\"“”#$%&\\'()\*+,\-\.\/:;<=>?@\[\\\]\^_`{|}~']", 'g')
  const analyzer = (token) => {
    let newToken = token.trim().replace(REGEXP_PATTERN_PUNCT, '').toLowerCase()

    if (eng.includes(newToken)) {
      return ''
    }

    return porterStemmer.stemmer(newToken)
  }

  // create a new index with a specific mapping
  const index = createIndex({
    title: 'text',
    body: 'text',
    topic: 'keyword'
  })

  // index some documents
  // raw documents are not stored in the index by default to optimize the index size
  // that's why we keep the data in a lookup mapping that can be used by the search to
  // get the documents later
  const docsLookup = {
    doc1: { title: 'Milk', body: 'A man is drinking milk.', topic: 'a' },
    doc2: { title: 'Bread', body: 'A man is eating breads.', topic: 'a' },
    doc3: { title: 'Butter', body: 'A man is eating bread and butter.', topic: 'b' }
  }
  const docsArray = Object.entries(docsLookup).map(([docId, doc]) => ({ _id: docId, ...doc }))

  docsArray.forEach((doc) => indexDocument(index, doc, analyzer))

  // make an example search on the 'body' and 'title' fields
  console.log(
    await searchIndex(
      index,
      'bread', {
        size: 10,
        queryFields: ['body', 'title'],
        filter: {
          topic: 'a'
        },
        getDocument: docId => docsLookup[docId]
      },
      analyzer,
      tokenizer
    )
  )
  // returns:
  // {
  //   total: 1,
  //   maxScore: 0.08530260953900706,
  //   hits: [ { _id: 'doc2', _score: 0.08530260953900706, _source: [Object] } ]
  // }
})()

See examples/.

API

`createIndex(mappings)`

TS Doc

Parameters

mappings: Mappings An object defining the fields of a document. Possible field types: text, keyword, number, date.

Return Value

Returns an index object to be used for querying and scoring. The raw documents are not included. Depending on the size of the text corpus, the size of the index can very.