Add self-closing MarkupEmpty element

[eemeli]

As discussed during this Monday's meeting, this PR adds the empty/standalone/self-closing MarkupEmpty element to the proposed spec, to represent elements that do not contain any body. This closes #238.

[stasm]

I agree with the change but I have comments about the formal implementation, and I'd like to see another iteration please.

[stasm]

This isn't LL(1) because it conflicts with the MarkupStart production. One way to fix this formally is something like the following (MarkupStart may need to be renamed):

MarkupStart ::= Name Option* '/'?
MarkupEnd ::= '/' Name

This should be OK because the EBNF only describes the grammar, not the data model. Assuming we end up with ElementStart, ElementEnd, and ElementEmpty in the data model, the parser can yield ElementStart when MarkupStart is missing the slash, and ElementEmpty when the slash is present.

[eemeli]

You're right. If keeping this syntax, it might make better to sense to have Markup and MarkupEnd in the syntax grammar, with Markup having a data model attribute like isEmpty: boolean determined by the presence (or not) of a trailing /.

The discussion in #269 is also relevant, as it may lead to different markup for Markup.

[stasm]

I realize that the lack of symmetry between Markup and MarkupEnd may be bothersome. How about a single Markup production?

Markup ::= Name Option* '/'? | '/' Name

[eemeli]

While that certainly technically works, it does rather look like two separate constructs crammed into one.

[stasm]

Right. EBNF requires a bit of formal accounting here and there to make it work, like the one I'm suggesting. Note that I only suggest to combine the grammar productions. The data model should have separate interfaces for all three element types.

[eemeli]

I updated the PR, merging MarkupStart and MarkupEmpty into one Markup element.

[eemeli]

Closing this, as the change proposed here has been obsoleted by subsequent conversation, esp. in #269.