Add test cases for spread attributes

[ForbesLindesay]

These currently already lex without any problems. I’d like to use these to replace &attributes and provide proper escaping in the process

[ForbesLindesay]

This only works if it's the first attribute in the list, and doesn't work for more complex JavaScript expressions. It would be nice to adapt it to work more reliably.

[TimothyGu]

This only works if it's the first attribute in the list, and doesn't work for more complex JavaScript expressions.

Yeah, since the dots might be part of a JavaScript expression (e.g. a(href=object .attr)). But since ... shouldn't be valid in normal JavaScript we can probably use a special case. Just thinking by talking.

[TimothyGu]

Just implemented that. All tests pass.

[TimothyGu]

Just a question: what is the foo? Why couldn't one just do (...foo)?

[ForbesLindesay]

My thinking was that it could mean you can namespace things: e.g.

div(...data=myDataAttributes)

compiled with:

{myDataAttributes: {foo: 'foo-val', bar: 'bar-val'}}

results in:

<div data-foo="foo-val" data-bar="bar-val"></div>

Probably this is the wrong syntax for that though. I think I would prefer to support something like:

div(data=...myDataAttributes)

or

div(data...=myDataAttributes)

as I think that reads better/more clearly. Probably best to remove support for having a "value" for spread attributes. If we make sure it's a syntax-error for now, we can always add it in later.

[ForbesLindesay]

Awesome. One last thing before we merge this: We should define a new token-type of spread-attr which should have mustEscape: true (we can think about syntax for un-escaped spread attributes some other time) and val: String

[TimothyGu]

So basically there are two classes of spread attributes:

1. a(...likeAndAttributes)
2. a(data=...specialSpreadAttributes)

I'll use spread-attribute for the first and spread-attribute-set for the second.

[ForbesLindesay]

Sounds good. I'm inclined to only implement spread-attribute for now, and leave spread-attribute-set as an idea we can investigate in the future. We'll need to remove the bonus test case 👍

[ForbesLindesay]

I'm labeling this as a breaking change because we currently allow attributes at the start to be like a(...foo='bar') and output <a ...foo="bar"></a>.

[TimothyGu]

Yep, sounds about right.

[TimothyGu]

Pushed a few commits that 1. added an unescaped version (div(...!unescapedAttrs)) and 2. also fixed the behavior of some regular attributes