Remove remaining jQuery from View tests

[akre54]

Now that #3003 is merged, we need an easy way for alternative View implementations to run against our View conformance test. This pulls jQuery out of the View tests except in for checking view.$el instanceof Backbone.$ (which all jQuery-like apis should support).

[eastridge]

@akre54 should Backbone.Nativeview provide a Backbone.$ to check against?

[akre54]

Why? It's not present and not needed.

[eastridge]

"which all jQuery-like apis should support" was what made me think of that, but I guess it's not really a jQuery API, but something that makes Backbone not need jQuery

[wyuenho]

We don't need the this in the end.

[akre54]

Good catch. Thanks.

[wyuenho]

Mind if I refactor the dispatchEvent shim and fix the custom event test?

[akre54]

Go for it. Not sure why Travis is throwing a fit. Maybe we have to expose CustomEvent from window?

[wyuenho]

Nah, non-standard custom events just need to be created in a special way.

https://developer.mozilla.org/en/docs/Web/API/CustomEvent

[braddunbar]

Take a look at #3077. I think that sort of thing will make this process much easier.

[akre54]

@braddunbar reverted View -> Backbone.View changes, and removed the setup method which was only being used to test the constructor. As long as the view implementation keeps a reference to Backbone.View (if needed), it should be fine to reassign to Backbone.View for testing purposes.

[wyuenho]

This looks good to me. It turns out there wasn't an easy way to fire a custom event on IE<=8.

Also, in case anyone is wondering why there's a click function, ariya/phantomjs#10795 is why.

[wyuenho]

Hey guys, can this be merged or is there something else you think we should do?

[braddunbar]

I do not think this one should be merged. From @jashkenas in #3077:

While I appreciate the sentiment, having a sanity test that Backbone View events are continuing to work properly with jQuery seems like a wise idea. It's not just about use testing the contract — it's also about being warned early if anything funny happens in future versions of jQuery (as has happened before).

I think we should be testing the real thing, not just that the arguments we pass are the arguments we pass.

Testing via mocked event creation is certainly not the real thing. Further, it somewhat obscures the contract.

[akre54]

Testing via mocked event creation is certainly not the real thing

How is this different from what we have now with $el.trigger('click')?

Tests aren't the place to catch issues with the latest jQuery. We've faced it before (with PATCH support, e.g.), and there almost always is a reasonable solution. Specs should never test your dependencies.

[braddunbar]

How is this different from what we have now with $el.trigger('click')?

jQuery#trigger is tested and used in real application code and a one-off implementation in our test suite isn't?

Tests aren't the place to catch issues with the latest jQuery. We've faced it before (with PATCH support, e.g.), and there almost always is a reasonable solution. Specs should never test your dependencies.

I'd prefer contract only tests in this case but using a trusted and tested library like jQuery is certainly better than an untested implementation written just for our tests.

[akre54]

Totally, and I can appreciate that. But this isn't some far-out custom code that we just pulled out of thin air. jQuery's code itself is more synthetic than what we're doing here since its on/trigger lifecycle lives entirely within the jQuery realm. Triggering a MouseEvent directly is much closer to what we're actually trying to test here (jquery itself uses elem.click() to trigger events bound outside of $.fn.on).

[akre54]

Hey @wyuenho, how come we couldn't just use element.click() directly here? It'd obviate this messiness here. Only downside I see is that the element would have to live in the DOM already.

[wyuenho]

ariya/phantomjs#10795

We can and should get rid of the click shim as soon as that phantom bug gets fixed. Either that or we use a different test runner like Karma that actually tests on real browsers.

[akre54]

Yeah it might be nice to get something like Sauce labs testing in here too since we're concerned about older browsers.

Shame about that bug though.

[akre54]

@braddunbar want to take one final pass through this and then either merge or close? I'm fine either way.

[wyuenho]

@akre54 Has PhantomJS 2 fixed this click event issue?