Change asap mutation observer implementation to touch text node

[josh]

Alternative PR to #1 as suggested by @azakus.

Opposed to setting attribute to the same value which may not trigger a mutation on buggy browsers or with a MutationObserver polyfill.

Hat tip to Polymer microtask.js
https://github.com/Polymer/platform-dev/blob/master/src/microtask.js

/cc @jakearchibald @azakus

[jakearchibald]

Looks good to me, but this polyfill is based on https://github.com/tildeio/rsvp.js/, and they'd benefit from the change too. I landed a commit over there a couple of days ago, once you land this I'll merge them in here and ship a new version.

Regarding the patch:

• I'm not sure what the browser support of the mutation polyfill is like, but textnode.data has better support than textnode.textContent (goes back to IE6, perhaps earlier) and it's fewer bytes
• I guess it's unlikely for iterations to get too big, but is (iterations = ++iterations % 2) better than iterations++?

I'll leave this open until I merge from rsvp

[josh]

tildeio/rsvp.js#170
tildeio/rsvp.js#171

[domenic]

-1. This is working as designed. When you use a MutationObserver polyfill, you explicitly opt in to flushing the queue yourself. If you don't flush the queue, then no promise handlers should fire. When you flush the queue, they should fire.

[domenic]

Wait. I didn't understand the change. So the MutationObserver polyfill is just buggy, and only requires explicit flush of the queue when you edit attributes? Well, I guess working around that bug is less bad. But why not submit a pull request to polymer to fix their polyfill to not trigger the bug in the first place?

[josh]

Its not practically unfixable since DOMAttrModified doesn't fire if the attribute is set to the same value. PR #1 suggested not trusting a non-native implementation in the first place.

[domenic]

Doesn't onpropertychange fire? Meh. I'm sure the polymer folks have already looked into this. Seems fine. Sorry for the initial negativity.

[stefanpenner]

@domenic can we confirm the polymer folks have looked into this? (I am not quite sure how they would work around this)

[josh]

tildeio/rsvp.js@3f75248

[josh]

@jakearchibald if you're just going to resync rspv.js source, I can close this.

[jakearchibald]

Changes went into rsvp.js, have been merged, new version (0.1.1) released with the changes \o/

@josh many thanks for this!

[stefanpenner]

Pretty cool, I didn't realize RSVP was used as the polyfil.

Keep me posted if you have any issues/ideas.

[jakearchibald]

Great isn't it? Looks like I did loads of work, but really it was just a few hours of deleting, renaming & making sure inheritance worked as spec'd.

(although I do point out the link in the first link of the readme, I'm not a total cretin)