This repository has been archived on 2023-08-20. You can view files and clone it, but cannot push or open issues or pull requests.
symfony/tests
Fabien Potencier bd5eda38fa merged branch Seldaek/eventdisp (PR #2572)
Commits
-------

fabe818 [EventDispatcher] Add reference to the EventDispatcher on the Event

Discussion
----------

[EventDispatcher] Add reference to the EventDispatcher on the Event

Bug fix: no
Feature addition: yes
Backwards compatibility break: no
Symfony2 tests pass: yes
Fixes the following tickets: -

I don't like registering event listeners unless they are really used, it seems wasteful. So I tend to register listeners for the response event in other listeners, only when they will be required. @stof has [brought to my attention](fb243ace83 (commitcomment-696467)) that this may cause issues in Silex or any other situation where event listeners are not lazy loaded, since it creates a circular reference in that case.

With this PR, avoiding the circular reference is possible, without bloating the response listener with unnecessary "do I need to do anything?" code.

---------------------------------------------------------------------------

by schmittjoh at 2011/11/06 05:28:39 -0800

Did you do any benchmarks? It's just a feeling, but registering a listener at runtime might be more expensive than just having it always executed.

Also, I find these dynamic listeners a bit of a code smell. They are not easily testable, and the control flow is harder to track. Besides, you do not take into account subrequests which might happen in between.

---------------------------------------------------------------------------

by Seldaek at 2011/11/06 09:34:27 -0800

I don't see why it would be slower, if it's a commonly fired event yes you blast away the `sorted` listener cache every time you add one, but most of the time those optional listeners are for the response, which is typically not sorted yet when you add the listener, so I don't think there is any overhead.

As for the code smell, of course it's a matter of preference, but I have the opposite view on control flow, I find it weird that listeners are registered when they are not used in the end, while doing it my way I think it's more clear what happens.

For sub-requests, I'm not sure what you mean. In this instance, and in most response listeners I have seen, the sub-requests are always ignored anyway by the listener.

---------------------------------------------------------------------------

by drak at 2011/11/10 06:07:45 -0800

I don't see how loading up the dispatcher with a bunch of callables can be expensive - it's just loading an array basically.

Wouldn't it be better to have a separate `DispatcherAwareEvent extends Event`

    class DispatcherAwareEvent extends Event
    {
        protected $dispatcher;

        public function setDispatcher(EventDispatcher $dispatcher)
        {
            $this->dispatcher = $dispatcher;
        }

        public function getDispatcher()
        {
            return $this->dispatcher;
        }

This can then be used as a base class for what you need `MyEvent extends DispatcherAwareEvent`

       $event = new MyEvent($dispatcher, $foo);
       $dispatcher->dispatch($event);

---------------------------------------------------------------------------

by Seldaek at 2011/11/10 06:18:57 -0800

@drak: Every event is part the event dispatching system and therefore should be aware of the dispatcher imo. It's not like the ContainerAwareInterface which is gluing things that do not especially have to know about the DIC together.

If we do that, then we have to start arguing every time we need the dispatcher in a given event, because the original author did not think it was necessary, and then that will only make it into the next minor version, etc. Not fun at all.

---------------------------------------------------------------------------

by drak at 2011/11/10 06:36:26 -0800

By the way, the event dispatchers looks to be pretty well optimized given the fact that it only sorts listeners if they are called, and then only once.

---------------------------------------------------------------------------

by drak at 2011/11/10 12:33:28 -0800

It just seems weird.  I mean, following on, why isn't the event name a compulsory parameter also? - again, you can say both ways, if you need it, make it part of your custom Event class, or since it's a required param to be able to dispatch an event in the first place, make it part of the base Event class.  All I'm saying it it seems suspicious when it could be achieved a different way.

For example, you could inject the dispatcher into the listener itself and then the event handler could access the dispatcher if it needs:

    class MyListener
    {
        public function __construct(EventDispatcher $eventDispatcher)
        {
            //...
        }

        public function someListener(Event $event)
        {
            //...
        }
    }

---------------------------------------------------------------------------

by stof at 2011/11/10 15:20:07 -0800

@drak The issue when injecting the dispatcher in the listener is described in the issue: circular dependency: you need to create the dispatcher before the listener (as it is injected in it) and when the listener is not lazy-loaded (in Silex for instance), you need to create it before the dispatcher.

---------------------------------------------------------------------------

by drak at 2011/11/10 21:15:45 -0800

Indeed, although it might not unreasonable to expect to create the dispatcher first...  but anyway I'm convinced!

Injecting the dispatcher could lead to some __very interesting possibilities__ as standard.  While we are at it though, we should have a getter and setter for `$name` in the `Event` class and `$event->setName($eventname)` in the `dispatch()` method.  Allowing an event to know it's name is very useful.  It allows a single listener to be registered for multiple names, and even makes the event object reusable. I don't know why $name was removed, it was in the Symfony 1 dispatcher and while the new dispatcher is brilliant from an OO point of view, missing the name as standard is a big shame.

+1 from me.
2011-11-22 09:32:53 +01:00
..
Symfony/Tests merged branch Seldaek/eventdisp (PR #2572) 2011-11-22 09:32:53 +01:00
bootstrap.php replaced symfony-project.org by symfony.com 2011-03-06 12:40:06 +01:00