[WIP]: ES6 Modules by leobalter · Pull Request #735 · qunitjs/qunit

leobalter · 2015-01-20T14:33:40Z

This is a WIP.

I've tried to set everything in ES6 modules, which showed to be more difficult than I thought with all the interdependencies in our code.

Currently, we need to make every module true independent, with that we can achieve the desired progress in this PR.

Also, the Browserify with 6to5 plugins aren't working as I expected. Bundling things still get then to use a require system, which I'm not aiming.

I'll follow @jzaefferer's suggestion and try http://jspm.io/

leobalter · 2015-01-20T14:35:58Z

Other problem: I still need to see how to handle JSCS to accept ES6. Probably easy, I just still need to go that further.

jzaefferer · 2015-01-20T14:46:28Z

I still need to see how to handle JSCS to accept ES6. Probably easy, I just still need to go that further.

We're not running jshint and jscs against src/ anyway, right? As long as the output in dist/ is more or less the same, there shouldn't be much of a difference.

Although, with proper modules and dependencies actually defined, we should also be able to validate src files directly...

jzaefferer · 2015-01-20T14:47:54Z

Let's stick with fixed version numbers, drop the ^.

wip. I'll fix that anyway.

leobalter · 2015-01-20T14:50:05Z

Although, with proper modules and dependencies actually defined, we should also be able to validate src files directly...

One of my main ideas on converting our modules into the es6 format is to allow running jshint and jscs against src/ files, are they're the ones we should test, not a built file.

Also, JSHint helped me a lot on fixing most of dependencies connections in this PR.

jzaefferer · 2015-01-20T14:50:56Z

Why is this import down here?

interdependency issue. Test need QUnit.config properly set. :(

jzaefferer · 2015-01-20T14:56:35Z

Well, now I see what you're talking about. Having modules depend on core, while those modules are also imported by core, is pretty bad. I think we need to, eventually, get rid of all these circular dependencies.

That said, overall it looks like we're on the right track. The import and export statements make sense and are easy enough to follow.

leobalter · 2015-01-20T15:08:36Z

this file got a huge diff because the indentation level reduced here. To ignore the whitespace change, see this diff including the ?w=1 parameter.

jzaefferer · 2015-02-04T20:14:16Z

Instead of jspm we should try to use esperanto, like we now do (or started doing) in PEP: jquery-archive/PEP#157

leobalter · 2015-02-06T17:14:51Z

This is getting more complex than expected and than I desired.

We have a lot of interdependencies that brings that complexity, I can make some inline comments to show it.

running grunt build will show some of the missing parts, but even if we solve the remaining, I don't like the interdependencies as they are right now. Solving them will look like a totally rewritten code, even if I don't try this.

At least, equiv, jsdiff and dump are easier to treat, as we still need to solve problems in test, assert, and the html reporter modules.

leobalter · 2015-02-06T17:17:12Z

we don't have the window as the global reference with these modules, it's necessary to fix it.

Does that only apply to window? What about setTimeout and other globals? How do you access the global scope correctly?

leobalter · 2015-02-06T18:30:22Z

I believe it's better stop this PR for now, and treat some circular dependencies that will allow us to turn everything into ES6 Modules with something not such aggressive in the changes.

The code isn't designed for modules, but I've identified a lot of good and bad things in this PR which I'll will help in future work.

Some of the good parts:

esperanto worked just fine
jsdiff, equiv and dump will be easy to handle
The grunt-esperanto module is getting some improvements to be able to used with UMD

Some parts still needed to improve:

The html reporter has some circular dependencies on core's config and urlParams. That shouldn't happen. Maybe the build should come from it, as it requires QUnit and attach the logging methods.
Test and Assert share a lot of cross dependencies with core, most of them can be identified to be detached from core.

[WIP]: ES6 Modules with Browserify and 6to5

307c215

jzaefferer reviewed Jan 20, 2015
View reviewed changes

[wip]: reduce diff by placing methods close to their original locations

85dfebe

leobalter added 2 commits January 20, 2015 09:57

[wip]: set strict version on dep packages

a2811f6

[wip]: reduce diff by moving methods [2]

b04380d

leobalter reviewed Jan 20, 2015
View reviewed changes

leobalter added 6 commits February 5, 2015 19:31

[fixup]: Add esperanto task config

59e7d37

[fixup]: 6to5 => Esperanto and update Grunt@0.4.5

53702f0

[fixup]: update grunt-jscs to 1.2.0

a3863f2

[fixup]: use PEP grunt task

5eb9929

HTML Reporter: Remove QUnit.init

5e89004

[fixup]: wip

f91fe2f

leobalter reviewed Feb 6, 2015
View reviewed changes

leobalter closed this Feb 6, 2015

Conversation

leobalter commented Jan 20, 2015

Uh oh!

leobalter commented Jan 20, 2015

Uh oh!

jzaefferer commented Jan 20, 2015

Uh oh!

jzaefferer Jan 20, 2015

Choose a reason for hiding this comment

Uh oh!

leobalter Jan 20, 2015

Choose a reason for hiding this comment

Uh oh!

leobalter commented Jan 20, 2015

Uh oh!

jzaefferer Jan 20, 2015

Choose a reason for hiding this comment

Uh oh!

leobalter Jan 20, 2015

Choose a reason for hiding this comment

Uh oh!

jzaefferer commented Jan 20, 2015

Uh oh!

leobalter Jan 20, 2015

Choose a reason for hiding this comment

Uh oh!

jzaefferer commented Feb 4, 2015

Uh oh!

leobalter commented Feb 6, 2015

Uh oh!

leobalter Feb 6, 2015

Choose a reason for hiding this comment

Uh oh!

jzaefferer Feb 6, 2015

Choose a reason for hiding this comment

Uh oh!

leobalter commented Feb 6, 2015

Uh oh!

Reviewers

Assignees

Labels

Milestone

Development

Uh oh!

3 participants