Exciting news!Announcing our $4.6M Series Seed. Learn more
Log in


Package Overview
File Explorer

Advanced tools


Powerful and versatile 100% javascript object database


Version published
Yearly downloads
increased by2.06%

Weekly downloads




Build Status

NDDB is a powerful and versatile object database for node.js and the browser.

NDDB (N-Dimensional DataBase) supports indexes, views, hashes, joins, group-by, basic statistics, custom operations, saving and loading from file system and browser localStorage, and much more.

Developer-friendly thanks to an easy api, detailed documentation, and wide test coverage.

List of features

  • Selecting: select, and, or
  • Sorting: sort, reverse, last, first, limit, distinct, shuffle
  • Indexing: view, index, hash, comparator
  • Custom callbacks: map, each, filter
  • Updating and Deletion: update, remove, clear
  • Advanced operations: split, join, concat, groupBy
  • Fetching and transformations: fetch, fetchArray, fetchKeyArray, fetchValues, fetchSubObj
  • Statistics operator: count, max, min, mean, stddev
  • Diff: diff, intersect
  • Skim: skim, keep
  • Iterator: previous, next, first, last
  • Tagging: tag
  • Event listener / emitter: on, off, emit
  • Saving and Loading: save, saveSync, load, loadSync, setWD, getWD, loadDir, loadDirSync


Load the library in Node.js:

const NDDB = require('NDDB'); // Backward-compatible mode. // const NDDB = require('NDDB').NDDB;

or in the browser add a script tag in the page:

<!-- Must load a version of NDDB that includes JSUS (see 'build/' dir) --> <script src="/path/to/nddb.js"></script>

Create an instance of NDDB:

let db = NDDB.db(); // let db = new NDDB(); // legacy

Insert an item into the database:

// Add one item to the database. db.insert({ painter: "Picasso", title: "Les Demoiselles d'Avignon", year: 1907 });

Import a collection of items:

let items = [ { painter: "Dali", title: "Portrait of Paul Eluard", year: 1929, portrait: true }, { painter: "Dali", title: "Barcelonese Mannequin", year: 1927 }, { painter: "Monet", title: "Water Lilies", year: 1906 }, { painter: "Monet", title: "Wheatstacks (End of Summer)", year: 1891 }, { painter: "Manet", title: "Olympia", year: 1863 } ]; // Import an array of items at once. db.importDB(items);

Retrieve the database size:

db.size(); // 6

Select Items

Select statements begin with select and can be refined with and and or statements. Select statements accept three input parameters:

  • 'property'
  • 'operator'
  • any additional number of arguments required by operator

Basic operators include standard logical operators:

  • =, ==, !=, >, >=, <, <=,

Advanced comparison operators include:

  • E: field exists (can be omitted, it is the default one)
  • ><: between values (expects an array as third parameter)
  • <>: not between values (expects an array as third parameter)
  • in: element is found in array (expects an array as third parameter)
  • !in: element is noi found in array (expects an array as third parameter)
  • LIKE: string SQL LIKE (case sensitive)
  • iLIKE: string SQL LIKE (case insensitive)

It is possible to access and compare nested properties simply separating them with ..

Select Examples

Select all paintings from Dali:

db.select('painter', '=', 'Dali'); // 2 items

Case sensitive LIKE operator:

db.select('painter', 'LIKE', 'M_net'); // 3 items

Select on multiple properties (*) with case insensitive LIKE:

db.select('*', 'iLIKE', '%e%'); // All items db.select(['painter', 'portrait'], 'iLIKE', '%e%') // 5 items

Select all portraits:

// Property 'portrait' must not be undefined. db.select('portrait'); // 1 item

Select all paintings from Dali that are before 1928:

db.select('painter', '=', 'Dali') .and('year', '<', 1928); // 1 item

Select all paintings of the beginning of XX's century:

db.select('year', '><', [1900, 1910]) // 2 items

Fetching items

Select statements are not evaluated until a fetch statement is invoked, returning the array of selected items, and preventing further chaining.

db.select('painter', '=', 'Dali').fetch(); // [ // { // painter: "Dali", // title: "Portrait of Paul Eluard", // year: 1929, // portrait: true // }, // { // painter: "Dali", // title: "Barcelonese Mannequin", // year: 1927 // } // ]

Other fetch methods can manipulate the items before they are returned.

// Create a new database without the items by Picasso. let newDb = db.select('painter', '!=', 'Picasso').breed(); // fetchValues // // Fetch all the values of specified properties and return them in an object. newDb.fetchValues(['painter', 'title']); // { // painter: [ 'Dali', 'Dali', 'Monet', 'Monet', 'Manet' ], // year: [ 1929, 1927, 1906, 1891, 1863 ] // } // fetchSubObj // // Keeps only specified properties in the objects, before returning them in // an array (items in the original database are NOT modified). newDb.fetchSubObj(['painter', 'title']); // [ // { // painter: "Dali", // year: 1929 // }, // { // painter: "Dali", // year: 1927 // }, // { // painter: "Monet", // year: 1906 // }, // { // painter: "Monet", // year: 1891 // }, // { // painter: "Manet", // year: 1863 // } // ] // fetchArray // // Returns the items as arrays. newDb.fetchArray() // [ // [ 'Dali', 'Portrait of Paul Eluard', 1929, true ], // [ 'Dali', 'Barcelonese Mannequin', 1927 ], // [ 'Monet', 'Water Lilies', 1906 ], // [ 'Monet', 'Wheatstacks (End of Summer)', 1891 ], // [ 'Manet', 'Olympia', 1863 ] // ] // fetchKeyArray // // Returns the items as arrays (including the keys). newDb.fetchKeyArray() // [ // [ // 'painter', 'Dali', 'title', 'Portrait of Paul Eluard', 'year', // 1929, 'portrait', true // ], // [ 'painter', 'Dali', 'title', 'Barcelonese Mannequin', 'year', 1927 ], // [ 'painter', 'Monet', 'title', 'Water Lilies', 'year', 1906 ], // [ // 'painter', 'Monet', 'title', 'Wheatstacks (End of Summer)', 'year', 1891 // ], // [ 'painter', 'Manet', 'title', 'Olympia', 'year', 1863 ] // ]


Define a global comparator function that sorts all the entries chronologically:

db.globalCompator = function (o1, o2) { if (o1.year < o2.year) return -1; if (o1.year > o2.year) return 1; return 0; };

Sort all the items (global comparator function is automatically used):

db.sort(); // Order: Manet, Monet, Monet, Picasso, Dali, Dali

Reverse the order of the items:

db.reverse(); // Order: Dali, Dali, Picasso, Monet, Monet, Manet

Define a custom comparator function for the name of the painter, which gives highest priorities to the canvases of Picasso:

db.compare('painter', function (o1, o2) { if (o1.painter === 'Picasso') return -1; if (o2.painter === 'Picasso') return 1; });

Sort all the paintings by painter using the new comparator:

db.sort('painter'); // Picasso is always listed first.


Splits the database in sub-databases, each containing semantically consistent set of entries:

// Let us add some cars to our previous database of paintings. let cars = [ { car: "Ferrari", model: "F10", speed: 350, }, { car: "Fiat", model: "500", speed: 100, }, { car: "BMW", model: "Z4", speed: 250, }, ]; // Default view: returns items with the value // of the property 'painter' !== undefined. db.view('painter'); // Make the view function explicit. db.view('art', function(o) { return o.painter; }); db.view('cars', function(o) { return o.car; }); db.rebuildIndexes(); db.size(); // 9 db.painter.size(); // NDDB with 6 art entries db.art.size(); // NDDB with 6 art entries db.cars.size(); // NDDB with 3 car entries


Define a custom hash function that creates a new view for each of the painters in the database:

db.hash('painter'); // Or the equivalent explicit function definition. db.hash('painter', function(o) { return o.painter; }); db.rebuildIndexes(); db.size(); // 6, unchanged; db.painter.Picasso; // NDDB with 1 element in db db.painter.Monet // NDDB with 2 elements in db db.painter.Manet // NDDB with 1 elements in db db.painter.Dali // NDDB with 2 elements in db

Listening to events

NDDB fires the following events: insert, update, remove, setwd, save, load. Users can listen to these events and modify their behavior.

Decorating objects on insert

Listen to the insert event and modify the inserted items by adding an index that is incremented sequentially:

let id = 0; function getMyId(){ return id++; }; db.on('insert', function(item) { item.myId = getMyId(); });

Canceling operations: insert, update, remove.

Event listeners can block the execution of the operation by returning false. No errors are thrown.

// Insert event. // Parameters: // - item: the item to insert. db.on('insert', function(item) { if (item.year > 3000) return false; // Item is not added. }); // Update event. // Parameters: // - item: the item to update. // - update: an object containing the properties to update/add. // - idx: the index of the item in the reference database (note: in a // sub-selection, the index of the item may differ from its index // in the main database.) db.on('update', function(item, update, idx) { if (update.year > 3000) return false; // Item is not updated. }); // Remove event. // Parameters: // - item: the item to remove. // - idx: the index of the item in the reference database (note: in a // sub-selection, the index of the item may differ from its index // in the main database.) db.on('remove', function(item, idx) { if (item.year < 3000) return false; // Item is not removed. });

Attention! The order in which the event listeners are added matters. If an event listener returns false, all successive event listeners are skipped.

Modifying save/load options.

// Save/load event (both sync or async). // Parameters: // - options: object with the user options for the save/load event. // - info: an object containing information about the save/load command, // which cannot be altered. Format: // { // file: 'path/to/file.csv', // format: 'csv', // cb: function() {}, // User defined function, if any. // } // db.on('save', function(options, info) { if (info.format === 'csv') { options.header = [ 'id', 'time', 'action']; // Modify header. } });

Intercept changes in working directory

// Set working directory event. // Parameters: // - wd: The new working directory. db.on('setwd', function(wd) { // Take note of the change, the value cannot be modified. });


Define a custom indexing function that gives fast, direct access to the items of the database;

db.index('id'); // Or the equivalent explicit function definition. db.index('id', function(o) { return o.id; }); db.rebuildIndexes(); db.id.get(0).name; // Picasso db.id.update(0, { comment: "Good job Pablo!" }); // Counts items in selection. db.select('comment').count(); // 1 let picasso = db.id.remove(0); db.size(); // (0) // Get all available keys in the index db.painter.getAllKeys(); // ['0','1', ... ] // Get all elements indexed by their key in one object db.painter.getAllKeyElements();

Default index

The property ._nddbid is added to every inserted item. The property is not enumerable (if the environment permits it), and all items are indexed against it:

db.nddbid.get('123456'); // Returns the item with nddbid equal to 123456.

Configuration Options

let logFunc = function(txt, level) { if (level > 0) { console.log(txt); } }; let options = { tags: {}, // Collection of tags update: { // On every insert, remove and update: indexes: true, // Updates the indexes, if any sort: true, // Sorts the items of the database pointer: true, // Moves the iterator to the last inserted element }, C: {}, // Collection of comparator functions H: {}, // Collection of hashing functions I: {}, // Collection of indexing functions V: {}, // Collection of view functions log: logFunc, // Default stdout logCtx: logCtx // The context of execution for the log function nddb_pointer: 4, // Set the pointer to element of index 4 globalCompare: function(o1, o2) { // Comparator. }, filters: { // Extends NDDB with new operators for select queries '%': function(d, value, comparator) { return function(elem) { if ((elem[d] % value) === 0) { return elem; } } } }, share: { // Contains objects that are copied by reference to // in every new instance of NDDB. sharedObj: sharedObj } } let nddb = NDDB.db(options); // or nddb = NDDB.db(); nddb.init(options);

Saving and Loading Items

The items in the database can be saved and loaded using the save and load methods, or their synchronous implementations saveSync and loadSync.

The methods loadDir and loadDirSync load an entire directory.

The following formats are available: csv, json, and ndjson.

Saving and loading to file system (node.js environment)

Two formats are natively supported: .json and .csv (automatically detected by the filename's extension. For unknown extensions, NDDB falls back to the default format (json, but it can be overridden).

It is possible to specify new formats using the addFormat method.

Save/Load Examples

// SAVING. // Saving items in JSON format. db.save('db.json', () => console.log("Saved db into 'db.json'") ); // Saving items in CSV format. db.save('db.csv', () => console.log("Saved db into db.csv'") ); // Saving items in CSV format. db.save('db.ndjson', () => console.log("Saved db into db.ndjson'") ); // Saving items synchronously in CSV format. db.saveSync('db.csv'); console.log("Saved db into db.csv'"); // Saving items in the default format (usually json). db.getDefaultFormat(); // json db.save('db.out', function() { console.log("Saved db into db.out'"); }); // Specifying the default format and saving into CSV. db.setDefaultFormat('csv'); db.save('db.out', function() { console.log("Saved db into db.out'"); }); // LOADING. // Loading items into database synchronously. db.loadSync('db.csv'); console.log("Loaded csv file into database"); // Loading 'adapted' items into database. db.load('db.csv', () => console.log("Loaded csv file into database") );

Loading an entire directory

The method loadDir and loadDirSync load an entire directory.

loadDir Options

In addition to the options of the native load method of the chosen format:

  • recursive: if TRUE, it will look into sub-directories. Default: FALSE.
  • maxRecLevel: the max level of recursion allowed. Default: 10.
  • filter: A filter function or a regex expression to apply to every file name.
  • dirFilter: A filter function or a regex expression to apply to every directory name.
  • onError: What to do in case of errors: 'continue' will skip the file with errors and go to the next one.
// Load files. let opts = { recursive: true, filter: 'bonus', // All files containing the word 'bonus'. dirFilter: (dir) => { return !~dir.indexOf("skip"); // Skip if directory contains word 'skip'. }; // Alternative filters: // filter: file => file === 'bonus.csv', // Only 'bonus.csv' files. // format: 'csv' // All 'csv' files. }; db.loadDirSync(DATADIR, opts);

Note: loadDir is not yet fully async. It loads files into the database asynchronously, but scans for files in the file system synchronously.

Adding a New Format

// Specify a new format. db.addFormat('asd', { save: function(db, file, cb, options) { // save file asynchronously. }, load: function(db, file, cb, options) { // load file asynchronously. }, saveSync: function(db, file, cb, options) { // save file synchronously. }, loadSync: function(db, file, cb, options) { // load file synchronously. } }); // Saving in the new format. db.save('db.asd');

Streaming items to file system (node.js environment)

The stream method automatically save items inserted into the database to the file system.

db.stream(); // Save items to [db name].[default format], for example: 'nddb.json'.

Stream options:

The stream method takes an optional configuration object:

  • format: the format: csv, json, ndbjson.
  • filename: path to file name (default [db name].[format])
  • delay: milliseconds to wait before copying items to file system (default 10)
  • journal: if TRUE, items are incapsulated in a data structure that contains information about the operation (insert, update, delete).

Journaling operations to file system (node.js environment)

The journal method keeps track of all operations (not just inserts).

db.journal(); // Save items to [db name].journal, for example: 'nddb.journal'.

This method is wrapper for the stream method with the journal flag TRUE.

Items are saved in ndjson format and can they imported in a new database with the importJournal method.

db.importJournal(); // All operations (inserts, updates, deletes) replayed.

CSV Advanced Options

Specifying an Adapter

// Transform items before saving them to CSV format. let options = { adapter: { // Double all numbers in column "A". A: function(item) { return item.A * 2; }, // Rename a property (must add shorterName to a custom header). shortName: 'muchLongerName' } }; db.save('db2.csv', options, () => { console.log("Saved db as csv into 'db2.csv'"); console.log("Numbers in column 'A' were doubled"); console.log("Values in column 'shortName' are taken from column 'muchLongerName'"); }); // Transform items before loading them into database. // Loading items into database. options = { adapter: { A: function(item) { return item.A / 2; } } }; db.load('db2.csv', options, () => { console.log("Loaded csv file into database"); console.log("Numbers in column 'A' were doubled"); });

Saving Updates Only

If you know already when a new set of items are added to the database, you can save incremental updates using the updatesOnly flag.

// Feedback view already created. db.comment.save('comments.csv', { // Custom header. header: [ 'timestamp', 'user', 'feedback' ], // Saves only updates from previous save command. updatesOnly: true });

Flatten Items

If a single user enters multiple items in the database, but you need only one row in the CSV file, you can use the flatten flag.

db.view('user').save('users.csv', { // Custom header. header: [ "user", "comment", "date", "name", "last", "rating" ], // Merges all items together. flatten: true, });

If you have multiple users in the database, the option flattenByGroup will create one CSV row per group (e.g., user).

db.user.save('users.csv', { // Custom header. header: [ "user", "comment", "date", "name", "last", "rating" ], // Merges all items together. flatten: true, // One row per user (can also be a function returning the id of the group). flattenByGroup: 'user', });

In case you need to periodically flatten the items, use the flatten option in combination with the updatesOnly flag.

Setting the current working directory (node.js environment)

It is possible to specify the current working directory to avoid typing long file paths.

// Saving items in JSON format. db.load('/home/this/user/on/that/dir/db.json'); db.load('/home/this/user/on/that/dir/db2.json'); db.load('/home/this/user/on/that/dir/db3.json'); // Can be shortened to: db.setWD('/home/this/user/on/that/dir'); db.load('db.json'); db.load('db2.json'); db.load('db3.json'); // Get current working directory: db.getWD(); // /home/this/user/on/that/dir/

List of all available options

{ flags: 'w', // The Node.js flag to write to fs. // Default: 'a' (append). encoding: 'utf-8', // The encoding of the file. mode: 0777, // The permission given to the file. // Default: 0666 // Options below are CSV ONLY: header: true, // Loading: // - true: use first line of // file as key names (default) // - false: use [ 'X1'...'XN' ] // as key names; // - array of strings: used as // is as key names; // - array of booleans: selects // key names in order from // columns in csv file // // Saving: // - true: use keys of first // item as column names (default) // - 'all': collect all keys // from all elements and use // as column names // - function: a callback that // takes each unique key in // the db and returns: // another substitute string, // an array of strings to add, // null to exclude the key, // undefined to keep it. // - false: no header // - array of strings: used as // is for column names (keys // not listed are omitted) adapter: { // Update the year property year: function(row) { // An object containing callbacks for return row['year']-1; // given csv column names. Callbacks take } // an object (a row of the csv file }, // file on load, or an item of the // database on save) and return a value to // be saved/loaded under that property name. separator: ',', // The character used as separator // between values. Default ','. quote: '"', // The character used as quote. // Default: '"'. escapeCharacter: '\\', // The char that should be skipped. // Default: \. (load only) lineBreak: '\n', // Line break character. Default: system's // default. bufferSize: 128 * 1024, // Number of bytes to read at once. // Default: 128 * 1024. // SAVE ONLY. bool2num: true, // If TRUE, booleans are converted to 0/1. na: 'NA', // Value for missing fields. Default: 'NA'. objectLevel: 2, // For saving only, the level of nested // objects to expand into csv columns flatten: true, // If TRUE, it flattens all items // currently selected into one row. flattenByGroup: 'player', // If set, there will one row per unique // value of desired group (here: 'player') updatesOnly: true, // If TRUE, saves only items that were // inserted into the database after // a file with the same name was last saved. updateDelay: 20000, // Number of milliseconds to wait before // saving updates. Default: 10000. }


NDDB relies on mocha and should.js for testing.

$ npm install # will load all necessary dependencies $ npm test # will run the test suite against nddb.js


Create your customized build of NDDB using the make file in the bin directory:

In order to run in the browser NDDB needs to have JSUS loaded. You can include it separately, or create a new build that includes it already. See the build help for options.

node make build // Standard build, node make build -a -o nddb-full // Full build

The build file will be created inside the build/ directory.






Subscribe to our newsletter

Get open source security insights delivered straight into your inbox. Be the first to learn about new features and product updates.

  • Terms
  • Privacy
  • Security

Made with ⚡️ by Socket Inc