How can jQuery deferred be used?

jQuery 1.5 brings the new Deferred object and the attached methods .when, .Deferred and ._Deferred.

For those who haven't used .Deferred before, I've annotated the source for it.

What are the possible usages of these new methods, how do we go about fitting them into patterns?

I have already read the API and the source, so I know what it does. My question is how can we use these new features in everyday code?

I have a simple example of a buffer class that calls AJAX requests in order. (Next one starts after the previous one finishes).

/* Class: Buffer
 *  methods: append
 *
 *  Constructor: takes a function which will be the task handler to be called
 *
 *  .append appends a task to the buffer. Buffer will only call a task when the 
 *  previous task has finished
 */
var Buffer = function(handler) {
    var tasks = [];
    // empty resolved deferred object
    var deferred = $.when();

    // handle the next object
    function handleNextTask() {
        // if the current deferred task has resolved and there are more tasks
        if (deferred.isResolved() && tasks.length > 0) {
            // grab a task
            var task = tasks.shift();
            // set the deferred to b开发者_JAVA百科e deferred returned from the handler
            deferred = handler(task);
            // if its not a deferred object then set it to be an empty deferred object
            if (!(deferred && deferred.promise)) {
                deferred = $.when();
            }
            // if we have tasks left then handle the next one when the current one 
            // is done.
            if (tasks.length > 0) {
                deferred.done(handleNextTask);
            }
        }
    }

    // appends a task.
    this.append = function(task) {
        // add to the array
        tasks.push(task);
        // handle the next task
        handleNextTask();
    };
};

I'm looking for demonstrations and possible uses of .Deferred and .when.

It would also be lovely to see examples of ._Deferred.

Linking to the new jQuery.ajax source for examples is cheating.

I am particularly interested in what techniques are available when we abstract away whether an operation is synchronously or asynchronously done.

---

The best use case I can think of is in caching AJAX responses. Here's a modified example from Rebecca Murphey's intro post on the topic:

var cache = {};

function getData( val ){

    // return either the cached value or jqXHR object wrapped Promise
    return $.when(
        cache[ val ] || 
        $.ajax('/foo/', {
            data: { value: val },
            dataType: 'json',
            success: function( resp ){
                cache[ val ] = resp;
            }
        })
    );
}

getData('foo').then(function(resp){
    // do something with the response, which may
    // or may not have been retrieved using an
    // XHR request.
});

Basically, if the value has already been requested once before it's returned immediately from the cache. Otherwise, an AJAX request fetches the data and adds it to the cache. The $.when/.then doesn't care about any of this; all you need to be concerned about is using the response, which is passed to the .then() handler in both cases. jQuery.when() handles a non-Promise/Deferred as a Completed one, immediately executing any .done() or .then() on the chain.

Deferreds are perfect for when the task may or may not operate asynchronously, and you want to abstract that condition out of the code.

Another real world example using the $.when helper:

$.when($.getJSON('/some/data/'), $.get('template.tpl')).then(function (data, tmpl) {

    $(tmpl) // create a jQuery object out of the template
    .tmpl(data) // compile it
    .appendTo("#target"); // insert it into the DOM

});

---

Here is a slightly different implementation of an AJAX cache as in ehynd's answer.

As noted in fortuneRice's follow-up question, ehynd's implementation didn't actually prevent multiple identical requests if the requests were performed before one of them had returned. That is,

for (var i=0; i<3; i++) {
    getData("xxx");
}

will most likely result in 3 AJAX requests if the result for "xxx" has not already been cached before.

This can be solved by caching the request's Deferreds instead of the result:

var cache = {};

function getData( val ){

    // Return a promise from the cache (if available)
    // or create a new one (a jqXHR object) and store it in the cache.
    var promise = cache[val];
    if (!promise) {
        promise = $.ajax('/foo/', {
            data: { value: val },
            dataType: 'json'
        });
        cache[val] = promise;
    }
    return promise;
}

$.when(getData('foo')).then(function(resp){
    // do something with the response, which may
    // or may not have been retreived using an
    // XHR request.
});

---

A deferred can be used in place of a mutex. This is essentially the same as the multiple ajax usage scenarios.

MUTEX

var mutex = 2;

setTimeout(function() {
 callback();
}, 800);

setTimeout(function() {
 callback();
}, 500);

function callback() {
 if (--mutex === 0) {
  //run code
 }
}

DEFERRED

function timeout(x) {
 var dfd = jQuery.Deferred();
 setTimeout(function() {
  dfd.resolve();
 }, x);
 return dfd.promise();
}

jQuery.when(
timeout(800), timeout(500)).done(function() {
 // run code
});

When using a Deferred as a mutex only, watch out for performance impacts (http://jsperf.com/deferred-vs-mutex/2). Though the convenience, as well as additional benefits supplied by a Deferred is well worth it, and in actual (user driven event based) usage the performance impact should not be noticeable.

---

This is a self-promotional answer, but I spent a few months researching this and presented the results at jQuery Conference San Francisco 2012.

Here is a free video of the talk:

https://www.youtube.com/watch?v=juRtEEsHI9E

---

Another use that I've been putting to good purpose is fetching data from multiple sources. In the example below, I'm fetching multiple, independent JSON schema objects used in an existing application for validation between a client and a REST server. In this case, I don't want the browser-side application to start loading data before it has all the schemas loaded. $.when.apply().then() is perfect for this. Thank to Raynos for pointers on using then(fn1, fn2) to monitor for error conditions.

fetch_sources = function (schema_urls) {
    var fetch_one = function (url) {
            return $.ajax({
                url: url,
                data: {},
                contentType: "application/json; charset=utf-8",
                dataType: "json",
            });
        }
    return $.map(schema_urls, fetch_one);
}

var promises = fetch_sources(data['schemas']);
$.when.apply(null, promises).then(

function () {
    var schemas = $.map(arguments, function (a) {
        return a[0]
    });
    start_application(schemas);
}, function () {
    console.log("FAIL", this, arguments);
});     

---

Another example using Deferreds to implement a cache for any kind of computation (typically some performance-intensive or long-running tasks):

var ResultsCache = function(computationFunction, cacheKeyGenerator) {
    this._cache = {};
    this._computationFunction = computationFunction;
    if (cacheKeyGenerator)
        this._cacheKeyGenerator = cacheKeyGenerator;
};

ResultsCache.prototype.compute = function() {
    // try to retrieve computation from cache
    var cacheKey = this._cacheKeyGenerator.apply(this, arguments);
    var promise = this._cache[cacheKey];

    // if not yet cached: start computation and store promise in cache 
    if (!promise) {
        var deferred = $.Deferred();
        promise = deferred.promise();
        this._cache[cacheKey] = promise;

        // perform the computation
        var args = Array.prototype.slice.call(arguments);
        args.push(deferred.resolve);
        this._computationFunction.apply(null, args);
    }

    return promise;
};

// Default cache key generator (works with Booleans, Strings, Numbers and Dates)
// You will need to create your own key generator if you work with Arrays etc.
ResultsCache.prototype._cacheKeyGenerator = function(args) {
    return Array.prototype.slice.call(arguments).join("|");
};

Here is an example of using this class to perform some (simulated heavy) calculation:

// The addingMachine will add two numbers
var addingMachine = new ResultsCache(function(a, b, resultHandler) {
    console.log("Performing computation: adding " + a + " and " + b);
    // simulate rather long calculation time by using a 1s timeout
    setTimeout(function() {
        var result = a + b;
        resultHandler(result);
    }, 1000);
});

addingMachine.compute(2, 4).then(function(result) {
    console.log("result: " + result);
});

addingMachine.compute(1, 1).then(function(result) {
    console.log("result: " + result);
});

// cached result will be used
addingMachine.compute(2, 4).then(function(result) {
    console.log("result: " + result);
});

The same underlying cache could be used to cache Ajax requests:

var ajaxCache = new ResultsCache(function(id, resultHandler) {
    console.log("Performing Ajax request for id '" + id + "'");
    $.getJSON('http://jsfiddle.net/echo/jsonp/?callback=?', {value: id}, function(data) {
        resultHandler(data.value);
    });
});

ajaxCache.compute("anID").then(function(result) {
    console.log("result: " + result);
});

ajaxCache.compute("anotherID").then(function(result) {
    console.log("result: " + result);
});

// cached result will be used
ajaxCache.compute("anID").then(function(result) {
    console.log("result: " + result);
});

You can play with the above code in this jsFiddle.

---

1) Use it to ensure an ordered execution of callbacks:

var step1 = new Deferred();
var step2 = new Deferred().done(function() { return step1 });
var step3 = new Deferred().done(function() { return step2 });

step1.done(function() { alert("Step 1") });
step2.done(function() { alert("Step 2") });
step3.done(function() { alert("All done") });
//now the 3 alerts will also be fired in order of 1,2,3
//no matter which Deferred gets resolved first.

step2.resolve();
step3.resolve();
step1.resolve();

2) Use it to verify the status of the app:

var loggedIn = logUserInNow(); //deferred
var databaseReady = openDatabaseNow(); //deferred

jQuery.when(loggedIn, databaseReady).then(function() {
  //do something
});

---

You can use a deferred object to make a fluid design that works well in webkit browsers. Webkit browsers will fire resize event for each pixel the window is resized, unlike FF and IE which fire the event only once for each resize. As a result, you have no control over the order in which the functions bound to your window resize event will execute. Something like this solves the problem:

var resizeQueue = new $.Deferred(); //new is optional but it sure is descriptive
resizeQueue.resolve();

function resizeAlgorithm() {
//some resize code here
}

$(window).resize(function() {
    resizeQueue.done(resizeAlgorithm);
});

This will serialize the execution of your code so that it executes as you intended it to. Beware of pitfalls when passing object methods as callbacks to a deferred. Once such method is executed as a callback to deferred, the 'this' reference will be overwritten with reference to the deferred object and will no longer refer to the object the method belongs to.

---

You can also integrate it with any 3rd-party libraries which makes use of JQuery.

One such library is Backbone, which is actually going to support Deferred in their next version.

---

I've just used Deferred in real code. In project jQuery Terminal I have function exec that call commands defined by user (like he was entering it and pressing enter), I've added Deferreds to the API and call exec with arrays. like this:

terminal.exec('command').then(function() {
   terminal.echo('command finished');
});

or

terminal.exec(['command 1', 'command 2', 'command 3']).then(function() {
   terminal.echo('all commands finished');
});

the commands can run async code, and exec need to call user code in order. My first api use pair of pause/resume calls and in new API I call those automatic when user return promise. So user code can just use

return $.get('/some/url');

or

var d = new $.Deferred();
setTimeout(function() {
    d.resolve("Hello Deferred"); // resolve value will be echoed
}, 500);
return d.promise();

I use code like this:

exec: function(command, silent, deferred) {
    var d;
    if ($.isArray(command)) {
        return $.when.apply($, $.map(command, function(command) {
            return self.exec(command, silent);
        }));
    }
    // both commands executed here (resume will call Term::exec)
    if (paused) {
        // delay command multiple time
        d = deferred || new $.Deferred();
        dalyed_commands.push([command, silent, d]);
        return d.promise();
    } else {
        // commands may return promise from user code
        // it will resolve exec promise when user promise
        // is resolved
        var ret = commands(command, silent, true, deferred);
        if (!ret) {
            if (deferred) {
                deferred.resolve(self);
                return deferred.promise();
            } else {
                d = new $.Deferred();
                ret = d.promise();
                ret.resolve();
            }
        }
        return ret;
    }
},

dalyed_commands is used in resume function that call exec again with all dalyed_commands.

and part of the commands function (I've stripped not related parts)

function commands(command, silent, exec, deferred) {

    var position = lines.length-1;
    // Call user interpreter function
    var result = interpreter.interpreter(command, self);
    // user code can return a promise
    if (result != undefined) {
        // new API - auto pause/resume when using promises
        self.pause();
        return $.when(result).then(function(result) {
            // don't echo result if user echo something
            if (result && position === lines.length-1) {
                display_object(result);
            }
            // resolve promise from exec. This will fire
            // code if used terminal::exec('command').then
            if (deferred) {
                deferred.resolve();
            }
            self.resume();
        });
    }
    // this is old API
    // if command call pause - wait until resume
    if (paused) {
        self.bind('resume.command', function() {
            // exec with resume/pause in user code
            if (deferred) {
                deferred.resolve();
            }
            self.unbind('resume.command');
        });
    } else {
        // this should not happen
        if (deferred) {
            deferred.resolve();
        }
    }
}

---

The answer by ehynds will not work, because it caches the responses data. It should cache the jqXHR which is also a Promise. Here is the correct code:

var cache = {};

function getData( val ){

    // return either the cached value or an
    // jqXHR object (which contains a promise)
    return cache[ val ] || $.ajax('/foo/', {
        data: { value: val },
        dataType: 'json',
        success: function(data, textStatus, jqXHR){
            cache[ val ] = jqXHR;
        }
    });
}

getData('foo').then(function(resp){
    // do something with the response, which may
    // or may not have been retreived using an
    // XHR request.
});

The answer by Julian D. will work correct and is a better solution.