OJ is a framework for writing web components and templates in frothy CoffeeScript or pure JavaScript. OJ provides a mechanism to rapidly build web applications using well encapsulated, modular code that doesn't rely on string templating or partially baked

npm install ojs
157 downloads in the last month

Welcome to OJ.

OJ(S) (OJ.js || OpenJS || Orange Julius || whatever you like) is a framework upon which to build web applications in pure (sometimes functional) JavaScript CoffeeScript.

OJ is packaged for Node and Bower. Other options abound.

  • Bower bower install --save ojs
  • From NPM last stable npm install --save ojs
  • Clone from Git git clone git@github.com:somecallmechief/oj.git


OJ is written in CoffeeScript. The distribution folder has the complete CS and JS for the project. The source folder includes CS, JS and source maps for each individual file. Feel free to use OJ in either flavor. While OJ probably doesn't have a lot of utility in the context of Node, I have made an effort to keep it Node-compatible--although there is absolutely no testing to this effect.


At its heart, OJ provides a simple API to build out the DOM:

div = OJ.body.div()
  .text "Aloha! Ahoy! Hola! Prevet!"
cell11 = div.table()
  .cell 1, 1
  .text 'Ahoy, column 1, row 1!'
  .cell 2, 1
  .text 'Aloha, column 1, row 2!'
span = cell11.span().text 'Aloha! Ahoy! Hola! Prevet!'

Nearly all standard nodes are accessible via chaining, starting with the body element. Every OJ node is chainable against the entire set of compatabile child nodes. For example:

div = OJ.body.div()
select = div.select()
  .button() #invalid child, will throw "TypeError: undefined is not a function"
  .option() #valid child of select

fieldSet = div.fieldset()

ol = div.ol()
  .span() #invalid child, will throw "TypeError: undefined is not a function"
  .li() #valid child of ol
  .div() #valid child of li
  .ul() #valid child of div
  .p() #invalid child, will throw "TypeError: undefined is not a function"
  .li() #valid child of li

This simple semantic makes it possible to build more complex web components, which are encapsualted as OJ classes.

div.grid() #creates a Bootstrap grid
  .tile 1, 1 #creates row 1, column 1
  .grid() #creates a nested grid on previous tile
  .tile 1, 1 #creates nested row, nested column

Components are currently a work in progress, but several functional components include jQuery Sparkline, jQuery Flotchart, and Bootstrap grids.


In order for OJ to scale, it has to be fast. By far, the fastest way to insert nodes into the DOM is by calling document.createElement and someElement.appendChild. Unfortunately, this is a tedious process to do manually. Rather than revert to anti-patterns like setting innerHTML, OJ leverages the speed of ThinDOM for light, fast, semantically accurate DOM insertion. This does come at a slight cost, but ThinDOM outperforms jQuery insertions by at least an order of magnitude--so it's a logical choice for the core of OJ's DOM manipulation.

While ThinDOM may be the fastest choice for getting your nodes into the DOM, it doesn't offer the cross-browser support and flexibility of jQuery's API--so for everything else you need to do to your nodes, OJ wraps around jQuery to provide access to all of the addClass and hide and on that you might need.


By design, OJ handles the generation of unique element IDs for every DOM node automatically under the hood. This provides faster lookups in the internal API and encourages you to reference nodes in memory as opposed to relying on "truth in DOM". For example, while you might write something like the following in jQuery:

jQuery '#myHolaDiv'
  .append '<div id="leavingDiv">Leaving...</div>'
jQuery '#myHolaDiv'
  .append '<div>Leaving......</div>'
jQuery '#myHolaDiv'
  .append '<div id="goneDiv">Gone</div>'

And at some point later fetch the node from the DOM:

leaving = jQuery '#leavingDiv'

In OJ, this is generally discouraged. Node lookups by ID are possible (if you maintain a reference to the ID OJ.nodes.get('myHolaDiv)), but it is usually better to consolidate behavior using a reference to the object and encapsulate the logic into components.

myHolaDiv = OJ.nodes.div();
leavingDiv = myHolaDiv.div()
  .text 'Leaving...'
goneDiv = myHolaDiv.div()
  .text 'Gone'


While frameworks like jQuery make it easy to inject anything the DOM, out of the box the framework does not necessarily encourage you to compose good HTML. In jQuery every node wrapper is a generic DOM node with the same sets of properties and methods as every other DOM node. In OJ, every node wrapper represents a specific type of DOM node (e.g. DIV, SPAN, TABLE) with a subset of the generic utility methods that you might get from the jQuery API but also a highly constrained, strongly validated superset of methods that are specific to the type of the node. For example, nothing (apart from reason) prevents one from writing this in jQuery:

table = jQuery '<table></table>'
div = jQuery '<div></div>'
table.append div

In OJ, the div method does not exist on the table class and thus will generate an exception:

table = OJ.nodes.table()
table.div() #TypeError: undefined is not a function

This does come at the cost of a learning curve, but the exceptions that OJ will generate for you are intended to ferret out mistakes that will never make their way into Production code. While still runtime exceptions, you will see these immediately--the first time you run the code or unit tests; and within a very short time will disappear completely.

One of OJ's implicit goals is to enable a style of development free from worrying about the minutia of the DOM in order to focus more on the UI/UX you want to deliver.

State of the Project

The Dream:

  • Basic HTML and CSS are fully encapsulated in strongly validated OJ classes.
  • Complex forms, layouts and composites are abstracted into thin OJ classes which are extensible with mixins.
  • Entire workflows are simplified into OJ structures which can be defined with just a few lines (or so) of JavaScript code.
  • Layouts are configurable simply by clicking and dragging the form. Validation and data binding come for free.

The Reality: Some of the most complex work has already been done. The SQL Builder demo is complete, minus actual AJAX requests. A fully fledged wrapper around IndexedDb is in place. Objects are queryable using SQL-like semantics. Raw DOM nodes are wrapped and ready. Much has been done, but much more still remains to do.

### TODO Priority 1 v0.2.0 Complete The core factory class is in place, but only SPAN and DIV have been implemented. TABLE is next, followed by INPUT (which will immediately spawn a factory subclass to handle input type: radio, checkbox, etc). SELECT (with child class OPTION) will be followed by the rest of the classic DOM Nodes.

As of v0.2.0, the entire library has been refactored to CoffeeScript. Anything non-essential to the DOM framework has been archived (the source is still included in the project under /src/archive). All standard nodes are supported: div, span, input, table, fieldset, p, b, br, ol, ul, li, select, option and more. Unit tests are in place and expanding to prove out all of the new nodes.

TODO Priority 2 v0.3.0

A secondary factory will need to implement components. There's still some implementation specing to do on how to distinguish between abstract components (e.g. an array of checkboxes) vs. concrete controls (e.g. an Address composite).

TODO Priority 3 v0.4.0

A FORM factory is on the table, so to speak. Additional specification work is needed to define a data structure which can be implicitly cast into an OJ form, generating an ordered layout of controls and to facilitate real time data.

Support or Contact

Feel like contributing? Wondering what in the sam hill I'm thinking? Challenge me to a duel; fork me; contact me; ignore me. Collaborate in real time [in my IDE of choice] (https://c9.io/somecallmechief/oj).


Everything I write is always Public Domain. Please take generously.

npm loves you