MarsDB is a lightweight client-side MongoDB-like database, Promise based, written in ES6

  • marsdb

Downloads in past


23130.6.118 years ago9 years agoMinified + gzip package size for marsdb in KB


Build Status npm version Coverage Status Dependency Status bitHound Overall Score Join the chat at GitHub stars
MarsDB is a lightweight client-side database. It's based on a Meteor's minimongo matching/modifying implementation. It's carefully written on ES6, have a Promise based interface and may be backed with any storage implementation (see plugins). It's also supports observable cursors.
MarsDB supports any kind of find/update/remove operations that Meteor's minimongo does. So, go to the Meteor docs for supported query/modifier operations.
You can use it in any JS environment (Browser, Electron, NW.js, Node.js).


  • Promise based API
  • Carefully written on ES6
  • Very very flexible – just take a look to the plugins section
  • Supports many of MongoDB query/modify operations – thanks to a Meteor's minimongo
  • Flexible pipeline – map, reduce, custom sorting function, filtering. All with a sexy JS interface (no ugly mongo's aggregation language)
  • Persistence API – all collections can be stored (and restored) with any kind of storage (in-memory, LocalStorage, LevelUP, etc)
  • Observable queries - live queries just like in Meteor, but with simplier interface
  • Reactive joins – out of the box



Meteor compatible client/server

Sometimes you can't use Meteor infrastructure. Maybe you need to build a custom client. Maybe you need to build a custom server with express and other modules. In meteor it can be done with a ton of hack. But the only reason why it's so ugly to do a simple things is because Meteor forces you to use their infrastructure. I'm trying to solve this issue with DDP client/server modules, based on MarsDB.


Using within non-ES6 environment

The ./dist folder contains already compiled to a ES5 code, but some polyfills needed. For using in a browser you must to include marsdb.polyfills.js before marsdb.min.js. In node.js you need to require('marsdb/polyfills'). It sets in a window/global: Promise, Set and Symbol.

Create a collection

import Collection from 'marsdb';
import LocalForageManager from 'marsdb-localforage';

// Default storage is in-memory
// Setup different storage managers
// (all documents will be save in a browser cache)

// Create collection wit new default storage
const users = new Collection('users');

Create an in-memory collection

import Collection from 'marsdb';
import LocalStorageManager from 'marsdb-localstorage';

// Set some defaults and create collection
const users = new Collection('users');

// But it may be useful to create in-memory
// collection without defined defaults
// (for example to save some session state)
const session = new Collection('session', {inMemory: true});

Find documents

const posts = new Collection('posts');
posts.find({author: 'Bob'})
  .project({author: 1})
  .then(docs => {
    // do something with docs

Find with pipeline (map, reduce, filter)

An order of pipeline methods invokation is important. Next pipeline operation gives as argument a result of a previous operation.
const posts = new Collection('posts');

// Get number of all comments in the DB
  .sortFunc((a, b) => a - b + 10)
  .filter(doc => Matsh.sqrt(doc.comment.length) > 1.5)
  .map(doc => doc.comments.length)
  .reduce((acum, val) => acum + val)
  .then(result => {
    // result is a number of all comments
    // in all found posts

// Result is `undefined` because posts
// is not exists and additional processing
// is not ran (thanks to `.ifNotEmpty()`)
posts.find({author: 'not_existing_name'})
  .aggregate(docs => docs[0])
  .aggregate(user =>

Find with observing changes

Observable cursor returned by a find and findOne methods of a collection. Updates of the cursor is batched and debounced (default batch size is 20 and debounce time is 1000 / 15 ms). You can change the paramters by batchSize and debounce methods of an observable cursor (methods is chained).
const posts = new Collection('posts');
const stopper = posts.find({tags: {$in: ['marsdb', 'is', 'awesome']}})
  .observe(docs => {
    // invoked on every result change
    // (on initial result too)
    stopper.stop(); // stops observing
  }).then(docs => {
    // invoked once on initial result
    // (after `observer` callback)

Find with joins

const users = new Collection('users');
const posts = new Collection('posts');
  .join(doc => {
    // Return a Promise for waiting of the result.
    return users.findOne(doc.authorId).then(user => {
      doc.authorObj = user;
      // any return is ignored
  .join(doc => {
    // For reactive join you must invoke `observe` instead `then`
    // That's it!
    return users.findOne(doc.authorId).observe(user => {
      doc.authorObj = user;
  .join((doc, updated) => {
    // Also any other “join” mutations supported
    doc.another = _cached_data_by_post[doc._id];

    // Manually update a joined parameter and propagate
    // update event from current cursor to a root
    // (`observe` callback invoked)
    setTimeout(() => {
      doc.another = 'some another user';
    }, 10);
  // Or just pass join spec object for fast joining
  // (only one `find` will be produced for all posts)
  .join({ authorId: users }) // posts[i].authorId will be user object
  .observe((posts) => {
    // do something with posts with authors
    // invoked any time when posts changed
    // (and when observed joins changed too)


const posts = new Collection('posts');
posts.insert({text: 'MarsDB is awesome'}).then(docId => {
  // Invoked after persisting document
  {text: 'MarsDB'},
  {text: 'is'},
  {text: 'awesome'}
).then(docsIds => {
  // invoked when all documents inserted


const posts = new Collection('posts');
  {authorId: {$in: [1, 2, 3]}},
  {$set: {text: 'noop'}}
).then(result => {
  console.log(result.modified) // count of modified docs
  console.log(result.updated) // array of updated docs
  console.log(result.original) // array of original docs

// Upsert (insert when nothing found)
  {authorId: "123"},
  {$set: {text: 'noop'}},
  {upsert: true}
).then(result => {
  // { authorId: "123", text: 'noop', _id: '...' }


const posts = new Collection('posts');
posts.remove({authorId: {$in: [1,2,3]}})
  .then(removedDocs => {
    // do something with removed documents array


  • Indexes support for some kind of simple requests {a: '^b'}, {a: {$lt: 9}}
  • Documentation


I'm waiting for your pull requests and issues. Don't forget to execute gulp lint before requesting. Accepted only requests without errors.


See License