Use the simpler evented-chain instead. It's got more focus, less features, and less surprises.

Create asynchronous callback chains. Make them powerful by chaining modules like async, and countless others that conform to the standard error-first callback and/or collection, iterator, consumer conventions.

Create a few functions that adhere to the error-first callback convention in node. Chain them together to create a new function.

var chain  = require('async-chains')

function cap (s)            { return s.toUpperCase()  }
function capEach (arr, cb)  { cb(null, arr.map(cap))  }
function toString (arr, cb) { cb(null, arr.join(' ')) }

var arrayToCapString = chain(capEach, toString)

arrayToCapString(process.argv.slice(2), function (err, result) {
    console.log(result)
})

When you invoke the new function, the arguments you invoke it with are passed to the first function in the chain. The callback arguments of that function are passed to the next function in the chain, and so on. If any function passes an error to a callback, the remainder of the steps in the chain are skipped, and the final callback receives the error.

async-chains was written to make expressing task sequences simpler, especially when leveraging modules like async, while maintaining the node callback pattern. Here's an example of a chain that leverages async to do some file ops:

var chain  = require('async-chains')
,   async  = require('async')
,   fs     = require('fs')
,   path   = require('path')

function readdirFullpath (dir, cb) {
    fs.readdir(dir, function (err, files) {
        cb(err, files.map(function (f) {
            return path.join(dir, f)
        }))
    })
}
function reFilter (pattern) {
    return function (str, cb) {
        cb(str.match(pattern) ? true : false)
    }
}
function fileStats (file, cb) {
    fs.stat(file, function (err, stats) {
        cb(err, {file: file, stats: stats})
    })
}
var getVimfileStats = chain(
    readdirFullpath, 
    chain.link.noError(async.filter, reFilter(/\.vim[^\/]*$/)), 
    chain.link(async.concat, fileStats))

getVimfileStats(process.env.HOME, function (err, result) {
    if (err) return console.error(err.message)
    result.forEach(function (item) {
        console.log('%s, last modified on %s', path.basename(item.file), item.stats.mtime)
    })
})

In this example, we use chain.link to interface with async functions. The async collection functions follow a format similar to many other node modules (collection, iterator, callback). The array and callback args will be supplied from the previous step in the chain, but in order to inject the iterator we want to use, we use chain.link(asyncCollectionFunction, iteratorFunction).

var chain = require('async-chains')

Create a new function (arg1, arg2, ..., finalCallback) that when invoked will have it's args passed to fn1 allong with a callback. When fn1 calls the supplied callback, the result of that callback is supplied as an argument to fn2, etc. If any callback is supplied an error argument, no further chain steps will be called, and the final callback will be immediately invoked with the error. Once the final function in the chain completes and calls it's callback, the arguments are passed to the finalCallback.

Returns a function usable in a chain that calls the collection function (collFn) given with any additional parameters needed (such is the case with the async limit variety, i.e. eachLimit), and iterator. The result of the previous chain step's output will be supplied as the first argument to the collection function.

This returns a function usable in a chain, as above, but should be used on async functions that do NOT return an error object. One such example is async.filter.

Create a function suitable for use as a standard node callback that feeds into a chain. Example:

var lastChange = chain(chain.link(async.concat, fs.stat), chain.link(async.reduce, 0, function (memo, item, cb) {
    cb(null, item.mtime > memo ? item.mtime : memo)
}))

fs.readdir(process.cwd(), chain.to(lastChange, function (err, result) {
    console.log(result)
}))

When a chain function is called, an event emitter is returned. This emitter emits step and complete events. None of these events are gauranteed. For example, a complete event is never emitted if an error is passed to the finalCallback. Nonetheless, these events may be useful for tracking progress. Example:

var chainFn = chain(fn1, fn2, fn3)
var progress = chainFn(function (err) {
    if (err) console.error(err)
})
progress
.on('step', function (step, args) {
    console.log('chainFn step %s executed', step)
})
.on('complete', function (args) {
    console.log('chainFn complete')
})

A step event is emitted for each function in the chain that is executed. Two arguments are passed to the listener, the first being the step index (0-based), and the second argument is an array of arguments that was passed to the function at that step.

A complete event is emitted once the chain function called the final callback. The listener of this event received a single argument, which is an array of arguments that was passed to the final callback.

npm test

Test scripts also support TAP when executed individually. You can use a tap consumer (such as node-tap) to consume all tests:

tap test

Test coverage reporting requires you to have istanbul installed.

npm install -g istanbul

Then you can run the coverage report:

npm run-script coverage

=============================== Coverage summary ===============================
Statements   : 100% ( 39/39 )
Branches     : 100% ( 8/8 )
Functions    : 100% ( 12/12 )
Lines        : 100% ( 39/39 )
================================================================================