NodeJS Streams with strong types via typescript, and helper methods to make dealing with object streams pleasent.

npm install barrage

import {Readable, Writable, Transform} from 'barrage';

These are native streams, but with a few extensions:

This is exactly like the built in source.pipe(destination, [options]) except that it also forwards any errors emitted by source to the destination. When your streams represent transformations, that is usually much more useful than the built in .pipe.

When the barrage is a readable stream, this method buffers the results and handles errors, resulting in a node.js style callback API. If there is no encoding parameter, the callback is called with an Array for the result. If encoding is 'buffer' then the callback is called with a single Buffer for the result. If any other string is passed as encoding, the encoding parameter is passed on to buffer.toString(encoding) and the result is therefore a String

This works like barrage.buffer, except that it does not buffer the result. It will wait for an end or finish event and then call the callback. If an error event is fired, the callback is called with that error. The callback is only ever called once.

If the callback parameter is absent, a Promises/A+ promise is returned instead.

This passes each chunk to transform and then pushes the result of calling transform to the output stream. You can either call this as a method on an existing barrage stream, or create a Transform stream by calling new barrage.Map

e.g.

function square() {
  return new barrage.Map(function (x) {
    return x * x
  })
}

It supports both being asynchronous, and parallel:

function load() {
  return new barrage.Map(function (stat, callback) {
    fs.readFile(stat.fullPath, callback)
  }, {parallel: 10})
}

When operating in parallel, the ordering of the resulting stream is always preserved.

It also supports promises

function load() {
  return new barrage.Map(function (stat) {
    return Promise.denodeify(fs.readFile)(stat.fullPath)
  }, {parallel: 10})
}

This is exactly like barrage.map / new barrage.Map except that transform should return true or false and the chunks will be filtered based on that value.

Take a function that maps an object onto an array or stream (or if for...of is supported by your version of node, any iterable), then return a stream for those individual items. e.g.

var source = new b.Readable({objectMode: true});
source._read = function () {
  this.push(1)
  this.push(2)
  this.push(3)
  this.push(null)
};
source.flatMap(function (x) {
  var source = new b.Readable({objectMode: true})
  source._read = function () {
    this.push(x * 1) // 1, 2, 3
    this.push(x * 2) // 2, 4, 6
    this.push(null)
  }
  return source
}).buffer().done(function (data) {
    assert.deepEqual(data, [1, 2, 2, 4, 3, 6])
    done()
  })

Takes a function that transforms a string and returns a Transform stream. e.g.

function coffeify(filename) {
  return new barrage.BufferTransform(function (src) {
    return compileCoffee(filename, src)
  }, 'utf8')
}
function compileCoffee(filename, src) {
  //do compilation and return a string
}
fs.createReadStream('src.coffee').pipe(coffeify('src.coffee')).pipe(fs.createWriteStream('src.js'))

This is mostly useful for processing files over stdio and creating browserify transforms.

The transform function may optionally take a callback argument (if it returns undefined) or return a promise (instead of a string).

MIT