CommonJS/RequireJS build tool
Pull request Compare This branch is 30 commits behind joewalker:master.
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Failed to load latest commit information.


DryIce is a CommonJS/RequireJS packaging tool for browser scripts, initially targeting the Skywriter/Ace project.

It is basically just a copy function. It takes input from a set of input files, which can be specified in various ways, optionally filters them and outputs them to something else.

DryIce is licensed under your choice of Mozilla Public License, GNU General Public License or GNU Lesser General Public License (MPL/GPL/LGPL tri-license).


RequireJS has a build tool which is nice and works well, but it requires Rhino and therefore Java. With DryIce, your whole build process can be in JavaScript.

DryIce produces a single file output that can include binary files (by base64 encoding them)

How to install DryIce

sudo npm install dryice

How does it work?

To copy a single file:

  source: 'foo.txt',
  dest: 'bar.txt'

To cat a bunch of files together:

  source: [ 'file1.js', 'file2.js' ],
  dest: 'output.js'

To cat together all the files in a directory:

  source: { root:'src' },
  dest: 'built.js'

As above, but only use the JavaScript files:

  source: { root:'src', include:/.*\.js$/ },
  dest: 'built.js'

As above, but exclude tests:

  source: { root:'src', include:/.*\.js$/: exclude:/test/ },
  dest: 'built.js'

If your set of files is very custom:

  source: function() {
    var files = [ 'file1.js' ];
    if (baz) files.push('file2.js');
    return files;
  dest: 'built.js'

We can filter the files on the way:

  source: /src/.*\.js$/,
  filter: copy.filter.uglify,
  dest: 'built.js'

This includes running multiple custom filters:

  source: 'src/index.html',
  filter: [
    function(data) {
      return data.replace(/Sun/, 'Oracle');
  dest: 'war/index.html'

Results can be stored and then used/reused:

var sources = copy.createDataObject();
  source: { root: 'src1' },
  dest: sources
  source: { root: 'src2' },
  dest: sources
  source: sources,
  dest: 'sources-uncompressed.js'
  source: sources,
  filter: copy.filter.uglify,
  dest: 'sources.js'

Data objects are just JS objects with a 'value' member, so you can do all sorts of things with them:

var test = copy.createDataObject();
  source: 'README.txt',
  dest: test


  source: { value: 'Hello, World!' },
  dest: 'basic.txt'

And you can mix and match your inputs:

  source: [
    { root: 'src', include: /.*\.js$/ },
    function() { return ''; }
  dest: 'mess.bin'

Common JS project dependency tracking:

var project = copy.createCommonJsProject({
    roots: [
    source: copy.source.commonjs({
        project: project,
        require: [ 'main', 'plugin/main' ]
    dest: ''

This digs around in the project source trees specified in the project for modules named in the 'require' statement. When it finds them it looks through them for require statements, and finds those, and so on.

Formal Parameter Description

The copy function takes a single parameter which is an object with 2 or 3 members: source, dest and optionally filter.


There are 6 ways to specify the input source(s)

  • A string is expected to point to a filename. At some stage we may allow them to point at directories too, however this can be achieved today using a find object (see below)

  • A find object points to a directory with 2 optional RegExps specifying what to exclude and include. e.g.

    { root: '/' } -> The entire filesystem { root: 'src', include: /.*.js$/ } -> All the JavaScript files in 'src' { root: 'src', exclude: /test/ } -> All non-test files under 'src'

  • A data object - something with a 'value' property. The implementation of copy.createDataObject() is simply return { value: '' };. We've batted around some ideas which involve making copy.createDataObject() smarter than it currently is, so it is advised to use this method rather than doing it yourself.

  • A based object. A based object is one with base and path members. They are roughly the same as the string baseObj.base + baseObj.path. Based objects are important when using CommonJS filters, because it tells the filter where the root of the hierarchy is, which lets us know the module name. For example:

    { base: '/etc', path:PATH } where BASE+PATH = filename

  • An array containing input source entries. The array does not have to be homogeneous.

  • A function which returns any input source entries.


The filter member is optional. If it exists, it should contain either a function or an array of functions. The function should have the following signature:

function filter(value, location) {
  return 'some string';

Where the parameters are as follows:

  • value. Either a string or a node Buffer. Most filters will work only with strings, so they should begin:

    if (typeof value !== 'string') {
        value = value.toString();

    Some filters will only work with Buffers (for example the base64 encoding filter) so they should begin:

    if (typeof value === 'string') {
        throw new Error('base64 filter needs to be the first in a filter set');

    At some stage we may allow filters to be marked up as to their requirements.

  • location. This will be (where possible) a based object or it could be a string if a based object is not available. It will be common to use one of the following idioms to work on a filename:

    if (location.base) {
        location = location.path;


    if (location.base) {
        location = location.base + location.path;

There are 2 points in a copy run where filters could be used, either before the individual sources are concatenated, or after. Some filters should be used in before (like common-js munging filters) and some afterwards (like compressors).

The default is to run filters after concatenation (when the location parameter will be undefined). To run filters before concatenation, the filter should be marked with onRead = true. For example:

function makeBlank(value, location) {
  return '';
makeBlank.onRead = true;

DryIce currently comes with 4 filters:

  • copy.filter.uglifyjs: Calls uglify on the input.
  • copy.filter.addDefines: Wraps the input to inline files fetched using RequireJSs text import feature.
  • copy.filter.base64: Similar to addDefines, but assumes the input is binary and should be base64 encoded.
  • copy.filter.moduleDefines: Replaces define lines to include the module name e.g. define(function(export, require, module) { ... }); is turned into define('module/name', function(export, require, module) { ... });


The dest property should be either a filename to which the output should be written (existing files will be over-written without warning), or a data object to which the data should be appended.

CommonJS Projects

CommonJS projects take a single object with the following properties:

  • roots: This is required. An array of directories that should be searched for your required modules and dependencies.

  • ignores: This is optional. An array of modules or dependencies that are required by your project that you would not like to be included in the build. For example, if you were making a build which did not need to support IE, you could do something like the following

          roots: [ '/path/to/project' ],
          ignores: [ 'dom/ie-compat', 'event/ie-compat' ]

    then wherever you had require('dom/ie-compat') or require('event/ie-compat') inside your build, undefined would be returned by require.

Where (is the project going)?

DryIce is useful in combining scripts for the browser, but it could also be used in a similar role on the server, we just need to enable 'pass through requires'.

There are some tweaks we'd still like to make to enable more filters and multiple destinations:

To recursively copy a directory:

copy({ source: 'foo', destDir: 'bar' });

To rename files as we copy them:

  source: { root:'src', include:/.*\.png$/ },
  destDir: { root:'built', replace:/png$/, with:'png.bak' }

To create a tarball (this is only missing the targz filter):

var version = copy.createDataObject();
copy({ source: 'VERSION.txt', dest: version });
  source: { root:'.' },
  filter: [ targz ],
  dest: 'scp://' + version + '.tar.gz'

I don't suppose you would ever actually want to do this, but in theory you could even do this:

  source: { root:'src', include:/.*\.java$/ },
  filter: javac,
  destDir: { root:'classes', replace:/java$/, with:'class' }

(Actually there would be issues with ordering that would make this hard, and Ant/Maven/etc is probably better. This is an illustration dammit!)