add define.exports

[stefanpenner]

• backfill some readme stuff

Opens the door to (template compiler and ember/glimmer need to learn about this next):

define.exports('my-app/templates/my-template.hbs', { hbs: true, "id": "VVZNWoRm", "block": "{\"statements\":[[1,[26,[\"welcome-page\"]],false],[0,\"\\n\"],[0,\"\\n\"],[1,[26,[\"outlet\"]],false]],\"locals\":[],\"named\":[],\"yields\":[],\"hasPartials\":false}", "meta": { "moduleName": "my-foo-app/templates/application.hbs" }});

instead of:

define("my-foo-app/templates/application", ["exports"], function (exports) {
  "use strict";

  Object.defineProperty(exports, "__esModule", {
    value: true
  });

  exports.default = Ember.HTMLBars.template({ "id": "VVZNWoRm", "block": "{\"statements\":[[1,[26,[\"welcome-page\"]],false],[0,\"\\n\"],[0,\"\\n\"],[1,[26,[\"outlet\"]],false]],\"locals\":[],\"named\":[],\"yields\":[],\"hasPartials\":false}", "meta": { "moduleName": "my-foo-app/templates/application.hbs" } });
});

benefits:

• no reification step
• no closure
• no juggling pre-parse semantics:

TODO:

• run tests on a larger code-base to get realistic #'s

[rwjblue]

Where does Ember.HTMLBars.template( bit go in the new semantics?

[stefanpenner]

Where does Ember.HTMLBars.template( bit go in the new semantics?

I would move it to the caller, which would detect require(templatePath).default.hbs === true or something.

[stefanpenner]

Some numbers (note the benchmarks only test runtime cost define + require, likely more tests are required re: eager vs lazy parse stuff and larger example \w many templates so we can see how gzip likes this)

Benchmark
  file: ./benchmarks/scenarios/define-many.js
  total:  73ms
  per op:  7.3ms

Benchmark
  file: ./benchmarks/scenarios/define-many-exports.js
  total:  47ms
  per op:  4.7ms

---

File size for a single small template (post gzip).

 t/size cat before.js | gzip -9 | wc -c
     318

 t/size cat after.js | gzip -9 | wc -c
     223

// before.js
define("my-foo-app/templates/application", ["exports"], function (exports) {
  "use strict";

  Object.defineProperty(exports, "__esModule", {
    value: true
  });

  exports.default = Ember.HTMLBars.template({ "id": "VVZNWoRm", "block": "{\"statements\":[[1,[26,[\"welcome-page\"]],false],[0,\"\\n\"],[0,\"\\n\"],[1,[26,[\"outlet\"]],false]],\"locals\":[],\"named\":[],\"yields\":[],\"hasPartials\":false}", "meta": { "moduleName": "my-foo-app/templates/application.hbs" } });
});

// after.js
define.exports('my-app/templates/my-template.hbs', { hbs: true, "id": "VVZNWoRm", "block": "{\"statements\":[[1,[26,[\"welcome-page\"]],false],[0,\"\\n\"],[0,\"\\n\"],[1,[26,[\"outlet\"]],false]],\"locals\":[],\"named\":[],\"yields\":[],\"hasPartials\":false}", "meta": { "moduleName": "my-foo-app/templates/application.hbs" }});

[rwjblue]

I would move it to the caller, which would detect require(templatePath).default.hbs === true or something.

Maybe, but if we do this, then it needs to include version info or something. Should be easy to put into the meta though...

[stefanpenner]

Maybe, but if we do this, then it needs to include version info or something. Should be easy to put into the meta though...

totally agree.

I'll run more thorough tests to confirm/deny my intuition (before we proceed with the required plumbing)

[rwjblue]

Another issue that we will have to address is how the templates get instantiated. In older Ember versions, templates were not instantiated but as of Ember 2.10 they are instantiated (and receive injections). Today, Ember.HTMLBars.template returns an object that adheres to the standard DI system contract (and we create instances of the templates by calling .create(injectionsHere) on them).

We could swap lookup to use factoryFor().class instead, but we just need to think it through...

[stefanpenner]

Today, Ember.HTMLBars.template returns an object that adheres to the standard DI system contract (and we create instances of the templates by calling .create(injectionsHere) on them).

Ya, I think we can wrap on the caller side for this as well.

[trentmwillis]

I like this!

[trentmwillis]

Missing a closing `

[trentmwillis]

Nit: Should be side-effects

[kratiahuja]

+1 to this.

[krisselden]

I'm in favor but this is going to require opt in from the caller side, one thing, the meta is extensible on the compilation side and present on the factory as well, so in the old and new world you can check the meta on the export.

[stefanpenner]

released as v4.5.0 🎉