Add deprecation for Route#renderTemplate

[patocallaghan]

Adds deprecation for Route#renderTemplate. See RFC 418 for more info.

I'm not going to lie, this was a bit rough with the amount of failing tests which needed to be fixed and then a few rebases in between 😅

id: 'route-render-template'
until: '4.0.0',
url: 'https://deprecations.emberjs.com/v3.x/#toc_route-render-template'
for: 'ember-source',
since: {
  enabled: '3.27.0',
},

[rwjblue]

Sorry this has been such a hassle @patocallaghan. I think we need to tweak the implementation a bit though. A large number of these tests don't seem to be using a custom renderTemplate, and they shouldn't be issuing a deprecation (if they do, that would mean that applications basically have no way to avoid the deprecation).

Basically, we need make it so that we avoid deprecation if you don't implement your own renderTemplate.

[rwjblue]

Hmm, I'm a little confused. Why is the deprecation being triggered in this test? I don't see it using renderTemplate...

[patocallaghan]

So yep this test does fail even though it doesn't have a custom renderTemplate.

It seems Route#renderTemplate is used internally by the framework in Route#setup. This must be why there were so many test failures 🤦‍♂️ I read the RFC but there was no mention of it so just assumed this was correct.

ember.js/packages/@ember/-internals/routing/lib/system/route.ts

Lines 990 to 992 in 28c18d0

	if (this._environment.options.shouldRender) {
	this.renderTemplate(controller, context);
	}

With the fact that the framework relies on this, I'm unsure how to proceed. How can we differentiate between Route#renderTemplate and a subclassed renderTemplate? Is it possible?

[rwjblue]

Yeah, there are a few different ways I think. Off the top of my head we could:

• Change the default implementation to use a private symbol, and only call the actual renderTemplate method when present
• Detect that the this.renderTemplate === Route.prototype.renderTemplate in the deprecate invocation

[patocallaghan]

Okay, thanks. I think the second approach is more straightforward so will give that a shot. For the first one could we still have a problem if people call super.renderTemplate(...arguments)?

[rwjblue]

Ya, agreed. So the first one is probably not feasible (sorry).

[patocallaghan]

@rwjblue I've updated with your proposed change and yep that made things much simpler. I should have asked for feedback earlier than just assumed all those test failures were accurate 🤦‍♂️

[rwjblue]

Thank you @patocallaghan!