refactor: use Pratt's parser for expressions in attributes

[PK1A]

This doesn't merge cleanly due to the custom attributes commit that landed but I'm opening a PR anyway for the early review.

@mlaval I will need your help with rebasing this one on top of your changes. Could you give me a hand with this one when you are available?

[PK1A]

This PR changes the syntax for the class attribute to the one discussed several weeks ago. After this PR is merged one would write class = "{{foo:bar}} {baz}" instead of class = "{foo:bar, baz}"

[benouat]

The changes were pretty easy to read, follow and most of it, understand. So I am really fine with it.

What I am not that keen of, is the amount of noise changes that you had to do to all these samples/*.txt files. To me it clearly shows that they are not that great as part of a testing environment. The coupling is definitely too strong.

But that's not related to this PR. Different topic, different discussion

[PK1A]

Yes, I agree that some of the tests are too low-level and thus pretty fragile. I think we will be gradually able to replace them with higher-level tests and / or more focused tests.

[PK1A]

@mlaval I've just pushed f733862 that introduces ability to get raw expression values when dealing with attributes. This should allow us to expose non-stringified values to custom attribute handlers.

Could we pair-program on the remaining part?

[PK1A]

@mlaval OK, so it seems like the class thing is sorted out but we seem to be getting issues from event handlers on IE :-( The error is really "cryptic" to me and I'm not sure why those errors are only triggered from the touch event handlers.... Could you give me a hand with this one?

[PK1A]

@ariatemplates/hashspace-contributors would be awesome if anyone could have another look at this PR - if not I'm going to land it as-is in order to not to block cleanups / syntax changes.

[benouat]

Latest commit 73d0c66 looks good to me