co

What is co?

The co npm package is a generator based flow-control utility for Node.js and the browser, making it easier to work with asynchronous JavaScript operations. It allows you to use generators to yield any function that returns a Promise. It can be used to simplify callback or promise-based code, especially in the context of async/await patterns.

What are co's main functionalities?

Sequential Execution

This feature allows for sequential execution of asynchronous tasks. The code sample demonstrates how you can use co to run promises in sequence using a generator function, which yields a promise that resolves to true.

co(function* () {
  var result = yield Promise.resolve(true);
  return result;
}).then(function (value) {
  console.log(value);
}, function (err) {
  console.error(err.stack);
});

Error Handling

This feature demonstrates how co can be used for error handling in asynchronous operations. The code sample shows a generator function yielding a promise that gets rejected, and the error is caught and logged.

co(function* () {
  try {
    yield Promise.reject(new Error('Oops!'));
  } catch (err) {
    console.error(err.message);
  }
});

Parallel Execution

This feature showcases how co can handle parallel execution of promises. The code sample illustrates a generator function yielding an array of promises, which co runs in parallel, and then logs the array of results.

co(function* () {
  var res = yield [
    Promise.resolve(1),
    Promise.resolve(2),
  ];
  return res;
}).then(function (value) {
  console.log(value);
});

Other packages similar to co

bluebird

Bluebird is a fully featured promise library with focus on innovative features and performance. It provides utilities for working with promises including but not limited to parallel execution, which is similar to what co offers. However, Bluebird does not use generator functions.

async

Async is a utility module which provides straight-forward, powerful functions for working with asynchronous JavaScript. Although it does not use promises or generators in the same way as co, it offers similar functionalities in terms of controlling the flow of asynchronous operations.

q

Q is a promise library for JavaScript which provides a toolset for creating and composing asynchronous promises. It is similar to co in that it helps manage asynchronous operations, but it does not utilize generator functions for flow control.

README

Co

Generator based flow-control goodness for nodejs (and soon the browser), using thunks or promises, letting you write non-blocking code in a nice-ish way.

Currently you must use the --harmony or --harmony-generators flags when running node 0.11.x to get access to generators.

Co is careful to relay any errors that occur back to the generator, including those within the thunk, or from the thunk's callback. "Uncaught" exceptions in the generator are then either passed co()'s thunk or thrown.

Installation

$ npm install co

Associated libraries

• co-fs - core fs function wrappers
• co-exec - core exec function wrapper

Example

var co = require('co');

co(function *(){
  var a = yield get('http://google.com');
  var b = yield get('http://yahoo.com');
  var c = yield get('http://cloudup.com');
  console.log(a.status);
  console.log(b.status);
  console.log(c.status);
})

Thunks vs promises

While co supports promises, you may return "thunks" from your functions, which otherwise behaves just like the traditional node-style callback with a signature of: (err, result).

For example take fs.readFile, we all know the signature is:

fs.readFile(path, encoding, function(err, result){

});

To work with Co we need a function to return another function of the same signature:

fs.readFile(path, encoding)(function(err, result){

});

Which basically looks like this:

function read(path, encoding) {
  return function(cb){
    fs.readFile(path, encoding, cb);
  }
}

This is what the co.wrap(fn) utility function does for you.

API

co(fn)

Pass a generator fn which is immediately invoked. Any yield expressions within must return a "thunk", at which point co() will defer execution.

var co = require('co');
var fs = require('fs');

function read(file) {
  return function(fn){
    fs.readFile(file, 'utf8', fn);
  }
}

co(function *(){
  var a = yield read('.gitignore');
  var b = yield read('Makefile');
  var c = yield read('package.json');
  console.log(a);
  console.log(b);
  console.log(c);
});

Nesting co() calls

The co() function itself returns a thunk, this means you can nest appropriately:

var co = require('co');
var fs = require('fs');

function size(file) {
  return function(fn){
    fs.stat(file, function(err, stat){
      if (err) return fn(err);
      fn(null, stat.size);
    });
  }
}

var foo = co(function *(){
  var a = yield size('.gitignore');
  var b = yield size('Makefile');
  var c = yield size('package.json');
  return [a, b, c];
})

var bar = co(function *(){
  var a = yield size('examples/parallel.js');
  var b = yield size('examples/nested.js');
  var c = yield size('examples/simple.js');
  return [a, b, c];
})

co(function *(){
  var a = yield foo;
  var b = yield bar;
  console.log(a);
  console.log(b);
});

Or a variation of the same thing:

var co = require('co');
var fs = require('fs');

function size(file) {
  return function(fn){
    fs.stat(file, function(err, stat){
      if (err) return fn(err);
      fn(null, stat.size);
    });
  }
}

co(function *(){
  var a = yield co(function *(){
    var a = yield size('.gitignore');
    var b = yield size('Makefile');
    var c = yield size('package.json');
    return [a, b, c];
  })

  var b = yield co(function *(){
    var a = yield size('examples/parallel.js');
    var b = yield size('examples/nested.js');
    var c = yield size('examples/simple.js');
    return [a, b, c];
  })

  console.log(a);
  console.log(b);
});

co() return values

Since co() returns a thunk, you may pass a function to this thunk to receive the return values from the generator. Any error that occurs is passed to this (sizes) function.

var co = require('co');
var fs = require('fs');

var read = co.wrap(fs.readFile);

var sizes = co(function *(){
  var a = yield read('.gitignore');
  var b = yield read('Makefile');
  var c = yield read('package.json');
  return [a.length, b.length, c.length];
});

sizes(function(err, res){
  console.log(res);
});

co.wrap(fn, [ctx])

The co.wrap() utility simply wraps a node-style function to return a thunk.

var co = require('co');
var fs = require('fs');

var read = co.wrap(fs.readFile);

co(function *(){
  var a = yield read('.gitignore');
  var b = yield read('Makefile', 'ascii');
  var c = yield read('package.json', 'utf8');
  console.log(a);
  console.log(b);
  console.log(c);
});

Optionally you may pass the fn's receiver as the ctx as shown here:

var co = require('co')
var redis = require('redis')
var db = redis.createClient()

db.set = co.wrap(db.set, db)
db.get = co.wrap(db.get, db)

co(function *(){
  yield db.set('foo', 'bar')
  yield db.set('bar', 'baz')

  var res = yield db.get('foo')
  console.log('foo -> %s', res);

  var res = yield db.get('bar')
  console.log('bar -> %s', res);
})

co.join(fn...)

The co.join() utility function allows you to pass multiple thunks, or an array of thunks and "join" them all into a single thunk which executes them all concurrently, instead of in sequence. Note that the resulting array ordering is retained.

var co = require('co');
var join = co.join;
var fs = require('fs');

function size(file) {
  return function(fn){
    fs.stat(file, function(err, stat){
      if (err) return fn(err);
      fn(null, stat.size);
    });
  }
}

co(function *(){
  var a = size('.gitignore');
  var b = size('index.js');
  var c = size('Makefile');
  var res = yield join(a, b, c);
  console.log(res);
  // => [ 13, 1687, 129 ]
});

As an alias of join(array) you may simply yield an array:

co(function *(){
  var a = size('.gitignore');
  var b = size('index.js');
  var c = size('Makefile');
  var res = yield [a, b, c];
  console.log(res);
  // => [ 13, 1687, 129 ]
});

FAQ

How does this compare to taskjs?

• it's smaller
• it's not a scheduler
• you can use thunks (functions)
• you can use promises

Does it work with streams?

Not out of the box, but if you're willing to fight node a bit you can wrestle streams into something usable:

co(function *(){
  var res = yield get('http://google.com');
  console.log('-> %s', res.status);

  var buf;
  while (buf = yield res.read()) {
    console.log(buf.toString());
  }

  console.log('done');
})

Performance

On my machine 30,000 sequential stat()s takes an avg of 570ms, while the same number of sequential stat()s with co() takes 610ms, aka the overhead introduced by generators is extremely negligable.

License

MIT