nerdypepper/sonar-tantivy

Search engine based on tantivy with a Node.js frontend

sonar

A tantivy based search engine for Node.js

Example

const Sonar = require('@archipel/sonar')

(async function () {
  const catalog = new Sonar('./data')
  const schema = getSchema()
  const index = await catalog.openOrCreate('index-name', schema)
  const docs = getDocs()
  await index.add(docs)
  const results = await index.query('world')
  console.log('query results', results)
})()

function getDocs () {
  return [
    { id: '0', title: 'Hello world!', body: 'tell me more' },
    { id: '1', title: 'Ola mundo!', body: 'que pasa pues' }
  ]
}

function getSchema () {
  return [
    {
      name: 'title',
      type: 'text',
      options: {
        indexing: { record: 'position', tokenizer: 'en_stem' },
        stored: true
      }
    },
    {
      name: 'body',
      type: 'text',
      options: {
        indexing: { record: 'position', tokenizer: 'en_stem' },
        stored: true
      }
    },
    {
      name: 'id',
      type: 'text',
      options: { indexing: null, stored: true }
    },
  ]
}

Installation

npm install @archipel/sonar

A postinstall script automatically tries to download a precompiled binary for the rust/tantivy part. If unsuccessfull the script will try to compile it if a rust toolchain is present.

API

const catalog = new Sonar(storage)

storage is a file system path where the index will be stored.

const index = await catalog.openOrCreate(indexName, schema)

indexName is a string to identifiy the index. It should only contain characters valid in file system paths. schema is the index schema, expressed as a JSON-serializable object following the tantivy schema definition. Documentation is not centralized atm, see example above.

const index = await catalog.create(indexName, schema, opts)

Create an index. Will throw if an index by this indexName exists already. opts are:

• ram: If true create an in-memory index

await index.add(docs)

docs is an array of documents with the same structure as the index schema.

const results = await index.query(query, [limit], [snippetField])

Query the index. At the moment only string queries are supported, see tantivy docs for details on the supported grammar. limit is the max number of documents to return (default 10). snippetField is the name of a field for which to return a result snippet with keywords highlighted (as HTML, with <b> tags)

const results = await catalog.multiQuery(query, indexes)

Query all indexes in the catalog. indexes is an array of index names.

To be expanded

Implementation details

The rust part is a wrapper around tantivy. It compiles to a binary. The binary is invoked with a storage path as only argument. It listens for newline-delimited JSON messages on STDIN, and replies in the same format.

The node part spawns the rust binary and communicates over the STDIO pipe. It adds a higher-level API around this simple RPC mechanism.

A npm postinstall step will try to download a precompiled binary of the rust part from Github releases. The binaries are compiled and deployed via Travis. If it cannot find a matching binary, it will try to compile if a rust toolchain is available. If the environment variable RUST_ENV=development is present, cargo run (without --release) will be invoked instead.