Port Yaku to core-js

[ysmood]

Continue after the #115. Sorry for so long. This time everything goes well. Only one file changed (92162b4), all the tests should pass.

When I debug the project, most of the unit test time is taken by the promise, it's unbearable for new contributors. Pulling the core code of promise out of the project will make the project more maintainable.

You don't have to run the promises-aplus-tests any more. Because as a separate project, yaku has already maintained the promises-aplus-tests on itself. So that you can focus on other parts of the project. Besides, I tested core-js against the promises-es6-tests, it didn't pass all the tests. Yaku did pass them all.

With Yaku, the performance will be better, the error handling will be less painful (such as long stack trace support), the gzipped size of the core.min.js won't change (27KB).

Here's the comparison:

name	unit tests	1ms async task	optional helpers	helpers	min js
yaku@0.11.6	✓	330ms / 106MB	✓	29	3.8KB
bluebird@3.3.1	x (18 failing)	265ms / 88MB	partial	100	52.2KB
es6-promise@3.1.2	x (52 failing)	426ms / 113MB	x	10	6.3KB
native@0.12.4	x ( 4 failing)	590ms / 173MB	x	13	0KB
core-js@2.1.0	x ( 4 failing)	838ms / 198MB	x	11	13.9KB
es6-shim@0.34.4	✓	950ms / 145MB	x	12	130.8KB
q@1.4.1	x (47 failing)	1599ms / 425MB	x	74	15.4KB

[zloirock]

Thanks for the PR, but I'm already not sure about adding dependencies to core-js. Some conceptions which I wanna use in the future makes it problematic. Need to think about it.

Without any tests I see some problems:

• Subclassing in .then should use @@species pattern, but uses this.constructor, it's completely incorrect.
• Iteration closing in Promise.all and Promise.race and some other problems with iterators.
• Static methods have separate feature detection and polyfilling. For example, .resolve should be always replaced in library version for some reasons.
• In environments without Symbol, well-known symbols @@iterator and @@species will not work in library version or if es6.symbol module will be required after es6.promise.
• Support rejectionHandled event.

I tested core-js against the promises-es6-tests, it didn't pass all the tests.

I know about it, it's a part of my not public Promise test case. core-js didn't pass 2 asserts because of 1 minor bug with the order of resolving. It's a minor problem, I was going to fix it later. Other 2 asserts can fail only with full core-js version because of Number iterator - use shim version for this test case.

What do you think about replacing core-js Promise core to modified Yaku Promise core instead of adding it as a dependency?

[ysmood]

• Subclassing in .then should use @@species pattern, but uses this.constructor, it's completely incorrect.

Fixed

• Iteration closing in Promise.all and Promise.race and some other problems with iterators.
• Static methods have separate feature detection and polyfilling. For example, .resolve should be always replaced in library version for some reasons.
• In environments without Symbol, well-known symbols @@iterator and @@species will not work in library version or if es6.symbol module will be required after es6.promise

Fixed

• Support rejectionHandled event.

Added

What do you think about replacing core-js Promise core to modified Yaku Promise core instead of adding it as a dependency?

No dependency now. I will merge the yaku.js to es6.promise.js later.

Any other problems?

[zloirock]

Something like that makes sense. I will explore it when I will have spare time.

[ysmood]

No harry, please take your time. I will add more test cases, like these:

https://github.com/ysmood/yaku/blob/master/test/basic.js

https://github.com/ysmood/yaku/blob/master/test/basic.js

[zloirock]

Closed because of inactivity.