Update autoprefixer to v2.0

[MoOx]

What do you think about that @ianstormtaylor ?

[MoOx]

I'll update that to 2.0 (ref #57)

[ianstormtaylor]

Curious to know about how much slower this makes Myth, we should have some big file sizes checks in the tests probably. What are the new features in 2.0 that we'd want?

For naming we can call the rework module itself Rework and keep the instance rework

[MoOx]

Just checkout the changelog https://github.com/ai/autoprefixer/blob/master/ChangeLog.md A lot has been done & it will be great to have that back. I'll run some bench before/after & use multiples concatened myth.io test case for that.

[MoOx]

So I just made some local test with materials from #83. Not sure how to handle that problem, since one day or another we will face the "outdated prefix stuff" issue :/

Before (using #83) github: 220ms myth.io: 15ms

After (using this branch & tests added in #83) github: 628ms myth.io: 46ms

3 times longer if we add autoprefixer 2 that use postcss.

• we parse & modify with rework
• we stringify for autoprefixer, that reparse & restringify
• we rereparse to allow myth to be used as a rework plugin

We should:

• add option to disable autoprefixer (some people might want to take care of that themselves), ref #77
• no serve a rework plugin by default (to avoid last rereparse), & so refactor to only serve a plugin if the user ask for it

What do you think guys ? poke @ai

[ai]

Unfortunately, Autoprefixer 2.0 is much smarter, that Autoprefixer 0.8. It groups properties by prefixes, works with spaces to visual cascade, etc.

[MoOx]

@ai did you check latest changes of @reworkcss/css ? Do you think we can find a way to make them interoperate with a map conversion ? I still don't get it why you create another css parser instead of contributing to this one :(

[ai]

@MoOx I think we need to use any postprocessors. If you will use only Rework processors it will become monopoly and development will be finished, as it was in Compass on Sass. There are a lot of new ideas ahead Rework and PostCSS. For example: xcss

But Rework team think to implement features, that Autoprefixer needs. But I think, it will anyway slowdown parser, so Myth benefits will be small.

[ai]

I have ideas to speed up parser by tokenizer (also it allows to use custom at-rules with any content), but it is only long terms plan.

[MoOx]

This thread is very interesting @ai. Thanks for sharing.

[MoOx]

So @ianstormtaylor what do you think about this ?

[ianstormtaylor]

Sounds good to me. Unfortunate still though, we might want to make our own prefixing rework plugin took get the speed back at some point, and would be nice to have it in Javascript anyways

[ai]

@ianstormtaylor there is a plan to rewrite Autoprefixer and PostCSS to ES6. But I afraid, that prefixes become to be unnecessary, when I finish :).

[MoOx]

Merged but I'll work asap on the ability to enable/disable features for people who cares about speed & control ( #77 ). I'll also remove the last reparsing by default & make it only if plugin is asked.

[ai]

Very nice! ^_^ waiting for release