On Github nolanlawson / nyc-camp-2015
db.get('my_doc_id').then(function(doc) { return db.remove(doc); }).then(function (result) { return db.info(); }).then(function (info) { // yay I'm done }).catch(function (err) { // ugh I got an error });I wrote it because I work on a very promise-heavy open-source library called PouchDB, and I was tired of perusing the questions on Stack Overflow and seeing the same errors again and again.
doSomething().then(function () { return doSomethingElse(); }); doSomething().then(function () { doSomethingElse(); }); doSomething().then(doSomethingElse()); doSomething().then(doSomethingElse);
doSomething().then(function () { // <- good pattern return doSomethingElse(); }); doSomething().then(function () { // <- trap! doSomethingElse(); }); doSomething().then(doSomethingElse()); // <- trap! doSomething().then(doSomethingElse); // <- good pattern
public String readFile(String filename) throws IOException { String output = ""; BufferedReader buff = new BufferedReader( new InputStreamReader(new FileInputStream(new File(filename)))); while (buff.ready()) { output += buff.readLine() + "\n"; } buff.close(); return output; }Here's an example from Java. Now, what's interesting here is that, from the point of view of a beginner, there's no difference between this part of the code, "readLine()" and the rest of tit. But if we were to promise this function without our larger code, it'd probably be the readLine() that ends up costing the most, especially if we're working with large files. In short, bottlenecks tend to pile up exactly where we're doing these sorts of InputStream operations. You might asay that the IOException communicates that, but there are plenty of parts of the Java core libraries that throw errors like this, e.g. the JSON parser or SAX parser, which doesn't require any I/O.
import requests; response = requests.get('http://oursite.com/ourapi/'); json = response.json() print jsonHere's an example from Python. Again, we really have no reason, as a beginner, to suspect that that second line will probably be much slower than any other line in our code. But of course an experienced developer will know that this HTTP server could take a very long time to respond, or not at all, so you better keep an eye on that part of the code.
var xhr = new XMLHttpRequest(); xhr.onerror = function (err) { console.log('whoops, got an error', err); }; xhr.onload = function (res) { console.log('woohoo, got a response', res); }; xhr.open('GET', 'http://oursite.com/ourapi/');The guy who designed JavaScript, Brendan Eich, was pretty smart. and one of the interesting design decisions he made for JavaScript was to make it single-threaded, with callbacks for AJAX requests. In JavaScript, this allows the single thread to continue working, e.g. responding to user clicks or updating the DOM, even when an HTTP request is ongoing.
$.ajax("http://oursite.com/ourapi/", { success: function (res) { console.log("woohoo, got a response", res); }, error: function (err) { console.log("whoops, got an error", err); } }); console.log('firing request');The guy who designed JavaScript, Brendan Eich, was pretty smart. and one of the interesting design decisions he made for JavaScript was to make it single-threaded, with callbacks for AJAX requests. In JavaScript, this allows the single thread to continue working, e.g. responding to user clicks or updating the DOM, even when an HTTP request is ongoing.
var fs = require('fs'); fs.readFile('/tmp/myfile.txt', 'utf-8', function (err, text) { if (err) { // handle error } else { // handle success } });This style continued with Node.js, and in fact it's the primary reason Ryan Dahl chose JavaScript for his single-threaded server framework. Here's a filesystem operation, which doesn't exist in the browser, but which looks very similar to the ajax request we saw earlier. So, in JavaScript, it's extremely obvious when a piece of code goes to the disk or to the network, because there's this awkward callback style you have to deal with. And it turns out this is kind of nice, because as this infographic shows, the network is much slower than the disk, which is much slower than anything done in-memory. My takeaway is: disk is about a zillion times slower than in-memory, and the network is about a zillion times slower than that. So JavaScript makes it really obvious when you're going to disk or the network, which can be nice for identifying bottlenecks or for ensuring that your single-threaded language doesn't spend any time waiting on I/O. And when webapps were using some light jQuery for a few ajax requests, which were almost never chained together, this was almost tolerable. But when Node.js came along, it just ended up being a huge hassle to write everything in this pyramid style when you just want to do a few I/O operations. The Node community tried to solve this early on by standardizing everything related to callbacks. But really this was just a convention, and nobody was forced to adhere to it, and people frequently messed it up. A common source of errors was accidentally calling a callback twice, or zero times, or with both an error and a result, etc. Lots of libraries sprang up proposing to solve this problem, such as async and q. But in fact they had very different and incompatible approaches, and jQuery and Angular had their own, so out of all this mess a few very smart people managed to hammer together a spec, and got almost everybody to agree on a small core of best practices, which they called the Promises A+ spec. Today this is enshrined in ES6, and it's in every major browser and soon Node.js, so it's here for the long haul. The only major holdouts are, sadly, jQuery and async. But you can convert their style to the official style, or just not use them. Okay, so everybody started using promises, and everything was great, right? Well, not exactly. First off, there is some confusion about which libraries and browsers actually use promises, and then second off there is confusion about how to use promises correctly. Let's talk about the libraries and browsers first.
doSomething(function (err) { if (err) { return handleError(err); } doSomethingElse(function (err) { if (err) { return handleError(err); } doAnotherThing(function (err) { if (err) { return handleError(err); } // handle success }); }); }); function handleError(err) { // handle error }
doSomething(function (err) { if (err) { return handleError(err); } // wait why is my code doSomethingElse(function (err) { if (err) { return handleError(err); } // on a deathmarch doAnotherThing(function (err) { if (err) { return handleError(err); } // to the right of the screen // handle success }); }); }); function handleError(err) { // handle error }
doSomething().then(function (result) { // handle success }).catch(function (err) { // handle error });
doSomething().then(function () { return doSomethingElse(); }).then(function () { return doAnotherThing(); }).then(function (result) { // yay, I'm done }).catch(function (err) { // boo, I got an error });
doSomething().then(function () { return doSomethingElse(); // ________ }).then(function () { // |__MEOW__| return doAnotherThing(); // \_/ }).then(function (result) { // /\_/\ // yay, I'm done // ____/ o o \ }).catch(function (err) { // /~____ =ø= / // boo, I got an error // (______)__m_m) });
doSomething().then(function () { doSomethingElse().then(function () { doAnotherThing().then(function () { // handle success }).catch(function (err) { // handle error }); }).catch(function (err) { // handle error }); }).catch(function (err) { // handle error });
doSomething().then(function () { return doSomethingElse(); }).then(function () { return doAnotherThing(); }).then(function (result) { // handle success }).catch(function (err) { // handle error });
playWithFire().then(function () { return liveDangerously(); }).then(function () { return betTheFarm(); }).then(function (result) { // handle success }); // ← forgot to catch. any errors will be swallowed
playWithFire().then(function () { return liveDangerously(); }).then(function () { return betTheFarm(); }).then(function (result) { // handle success }).catch(console.log.bind(console)); // ← this is badass
goToTheATM().then(function () { grabMyCash(); }).then(function () { grabMyCard(); }).then(function (result) { // grabMyCash() and grabMyCard() // are not done yet! }).catch(console.log.bind(console));
goToTheATM().then(function () { return grabMyCash(); }).then(function () { return grabMyCard(); }).then(function (result) { // yay, everything is done }).catch(console.log.bind(console));
someLibrary.doSomething().then(function () { // I'm inside a then() function! });
someLibrary.doSomething().then(function () { // 1) return another promise return someLibrary.doSomethingElse(); // 2) return a synchronous value return {hooray: true}; // 3) throw a synchronous error throw new Error('oh noes'); });
db.get('user:nolan').then(function (user) { if (user.isLoggedOut()) { throw new Error('user logged out!'); // throwing a synchronous error! } if (inMemoryCache[user.id]) { return inMemoryCache[user.id]; // returning a synchronous value! } return db.get('account:' + user.id); // returning a promise! }).then(function (userAccount) { // I got a user account! }).catch(function (err) { // Boo, I got an error! });
var bluebird = require('bluebird'); var fs = bluebird.promisifyAll(require('fs')); fs.readFileAsync('/tmp/myfile.txt', 'utf-8').then(function (text) { // handle success }).catch(function (err) { // handle error });
var Promise = require('some-valid-library'); var promise = new Promise(function (resolve, reject) { // roll your own promise inside here }); promise.then(); promise.catch(); Promise.resolve(); Promise.reject(); Promise.all(); Promise.race();
Promise.resolve().then(function () { throw new Error("aw shucks"); });
var db = new PouchDB('mydb'); db.post({}).then(function (result) { // post a new doc return db.get(result.id); // fetch the doc }).then(function (doc) { console.log(doc); // log the doc }).catch(function (err) { console.log(err); // log any errors });
let db = new PouchDB('mydb'); try { let result = await db.post({}); let doc = await db.get(result.id); console.log(doc); } catch (err) { console.log(err); }