Once

[tbranyen]

once method

[jashkenas]

I'm afraid I still don't think this is a common enough use case to bake into Backbone.Events ... but you can use _.once to accomplish the same thing, when you need it.

[powmedia]

+1 for a proper .once() method which removes the event listener; could help prevent memory leaks

[TheCloudlessSky]

+1 for this. @jashkenas - it's not the same thing since _.once would still keep the handler bound after the first time it's called. Here's my use case:

1. Create a model.
2. Bind once to the sync event to the model.
3. Show a dialog and when closed, save the model.
4. The once'd sync event is triggered (with { wait: true}) and the model is added to the collection.

For example:

var newAppointment = new mos.model.Appointment();
newAppointment.once('sync', function() {
  // Only add to the collection after it has been saved.
  collection.add(newAppointment);
});

var dialog = new mos.ui.AppointmentDialog({
  model: newAppointment,
  mode: mos.ui.DialogMode.CREATE
});

dialog.render(); // User then clicks 'Save' which would trigger "model.save()"

[nibblebot]

+1 @jashkenas please re-open this for consideration. _.once does not provide the same functionality

[jashkenas]

I would, but GitHub seems to have disappeared the "reopen" button that used to be on these pages. If the button comes back, I'd be glad to reopen it.

[tbranyen]

I re-opened this issue and rebased to merge cleanly. A few other things as well: I updated the source code to not modify the original callback that was passed in. This previous method didn't sit well with me, because it meant code like this wouldn't work:

var obj = _.extend({}, Backbone.Events);
var test = function() { console.log("triggered"); };
obj.once("test", test);

// Would not work.
obj.off("test", test);

This new method simply binds an additional "cleanup" event that removes immediately after the first event fires. @jashkenas @braddunbar code review of this would be awesome.

I also updated the qunit tests to use equal instead of equals all passing atm.

[gsamokovarov]

I like this one too, use it quite regularly. +1

[braddunbar]

I believe the above should be this.on(events, unbind, this).

var context = {};
obj.on('event', function(){}, context);
// Dies because `unbind` is called with `context` instead of `obj`.
obj.trigger('event');

Also, I think the convention would be to return this, though I'm tempted to use the simpler form and return unbind for offing the handler.

[braddunbar]

No need to default context here since it's done for us in trigger.

[braddunbar]

Mornin' @tbranyen! The approach is interesting but I have to wonder, given the way once is essentially a callback, how often it would be useful to off the handler anyway. It's a good deal of overhead for something that can already be done with little effort and no extra cost.

model.on('event', function handler() {
  model.off('event', handler);
  // ...
}, model);

// ...

if (condition) model.off('event', handler);

[tbranyen]

@braddunbar Is it a common pattern to use a NFE inside of event binding to emulate once? Personally I haven't seen this, but a cursory glance at EventEmitter, EventEmitter2, Emitter, and jQuery all have a once implementation.

Also: Ember.js and Spine.js

[braddunbar]

Good question. I only used the NFE because it was easier to type out in a comment.

[MFoster]

@tbranyen +1. It's very convenient for setting off one time operations that need to respond once to a recurring event, such as a set up operation.

[nibblebot]

@tbranyen Interesting. I hadn't considered the use of case of .off'ing a .once. I can see where you might want this flexibility but now we have 2x overhead for .once (2 .on handlers). I suppose I'd be willing to make that trade though :)

[tbranyen]

@nibblebot They are synchronous calls, the overhead isn't as bad at what you'd think. It's basically an additional function call which we were already doing.

[tbranyen]

@braddunbar I changed the implementation around, does this meet your concerns? There is almost no overhead with this new approach.

[braddunbar]

@tbranyen Actually, I prefer the previous version. I think it's rather elegant and I'm ok with the extra call to off, especially since it's still possible to do it yourself if you like.

I know the extra cost to trigger is small, but I think that any changes we make to trigger should be to make it faster. It also doesn't handle obj.once('all', ...), which would add another check.

Thanks for the discussion and my apologies for the back and forth.

[tgriesser]

@braddunbar did we determine if there'd be any potential issues with the NFE?

[braddunbar]

I've looked through the NFE bugs and I don't see any potential issues, but I wouldn't mind if we broke it out just to be safe.

[jdalton]

NFE leak in IE < 9 because it creates 2 distinct function; 1 function declaration and 1 function expression.
See http://kangax.github.com/nfe/#jscript-memory-management

[philfreo]

👍 for once

[tbranyen]

Just found another place in my code in which once would be heavily desired.

// Special cases for when a parent View that has not been rendered is
// involved.
if (manager.parent && !manager.parent.__manager__.hasRendered) {
  return manager.parent.on("afterRender", function() {
    // Unbind this event... really wish we had once.
    manager.parent.off(null, null, this);

    // Trigger the afterRender and set hasRendered.
    completeRender();
  }, this);
}

[tbranyen]

Another situation where it would be much cleaner to have once:

        ctor.prototype.fetch = function(options) {
            // Set the nSync property to true indicating we are doing something.
            this.nSync = true;

            function reallyWantOnce() {
                this.off("reset", reallyWantOnce);
                this.nSync = false;
            }

            this.on("reset", reallyWantOnce, this);

[khepin]

Definitely would be interesting for me like I already said in #594

[caseywebdev]

@tbranyen I sent you a PR (tbranyen#1) that accounts for once in the callback, avoiding a potential infinite loop. Test included too.

[jergason]

Just another word of support: I would love the once functionality in Backbone events. It is in the node EventEmitter, and is very useful there.

[BlakeWilliams]

+1

[grydstedt]

+1

[tbranyen]

Anything I can do to get this merged? @jashkenas @braddunbar ?

[caseywebdev]

@tbranyen please take a look at my pull request on your branch. It prevents an easily avoidable more-than-once "once" case.

[tbranyen]

Awesome, updated @caseywebdev 👯

[caseywebdev]

No prob! I love once =D 💃

[braddunbar]

@tbranyen I posted some comments above, though I'll leave the merging to @jashkenas.

@tbranyen Actually, I prefer the previous version. I think it's rather elegant and I'm ok with the extra call to off, especially since it's still possible to do it yourself if you like.

I know the extra cost to trigger is small, but I think that any changes we make to trigger should be to make it faster. It also doesn't handle obj.once('all', ...), which would add another check.

Thanks for the discussion and my apologies for the back and forth.

Another reason for the previous version, is that the current version fails in the following instance:

var a = new Backbone.Model().once('event', f);
var b = new Backbone.Model().on('event', f);

a.trigger('event'); // Calls `f`
a.trigger('event'); // Noop

b.trigger('event'); // Calls `f`
b.trigger('event'); // Should call f, but doesn't

I submitted a pull request with a failing test case.

[RStankov]

+1

[tbranyen]

@jashkenas @braddunbar Let me know if you have any further concerns before we merge it.

[braddunbar]

Removing all callbacks for the event isn't appropriate here. The other events aren't necessarily once callbacks. Pull request with failing test case incoming.

[tgriesser]

leftover debugger

[braddunbar]

snip

[tgriesser]

+1 - I've run into several cases where this would have been helpful to have.

[jashkenas]

Alright folks -- I've just pushed a patch that implements "once". Please take a very close look at it -- I think it's an elegant little implementation -- and add any additional test cases that you think I might be missing.