Do we need to parse the grammar at all in the Word converter?

[jskeet]

C# code in the standard isn't parsed - it's just colorized.

The ANTLR grammar, however, is parsed. I suspect that's for historical reasons, back when the Word converter would also gather all the grammar rules together and perform a certain amount of validation.

I'm wondering whether we could fix #394 (and simplify the code!) just by not parsing the ANTLR at all. It might be tricky to get it colorized that way, given that Colorize doesn't have any built-in support, but I suspect that "leave it as formatted in Markdown, but as if it's plain text" is actually better than trying to parse and handle it ourselves.

Will give that a try...

[jskeet]

Looks like we also use the parsed grammar to link to bookmarks. We could potentially still just use the plain text and keep parsing it to get the production names... or lose the bookmarks. Will prototype both ways.

[jskeet]

Ah - no, we only use the parsed grammar production if the language is "antlr" rather than "ANTLR", which it never is. (See #342.)

So we really don't lose anything by not parsing it at all, given that we've got a separate validator. Let's go with that simple option.