Comment parsing priority

[fennifith]

Hello! I've been spending a bit of time writing an (incomplete) Kotlin implementation of KDL (with serialization support), and I really like how clear/readable it is to work with!

I do have a slight concern about the way that comments are supposed to be handled - I believe the current spec prioritizes identifiers over the comment syntax so that /*identifier (unless a corresponding */ appears later in the file), ident/**/ifier, identi//fier (etc.) are all valid nodes (please correct me if I am mistaken!).

Currently, my tokenizer always prioritizes the single/multiline comment syntax at any point except during a string. When processing an identifier, the program isn't sure whether it is a value/keyword/identifier until it finds the end, at which point it would be less efficient to backtrace and remove a comment block that has just been discovered (i.e. in the first example above). I would like to avoid doing that if possible, as I'm trying to enable partial parsing of large files - so that nodes can be skipped or functions can do an early return before the entire document has been read.

Additionally, from a user perspective, I feel like this could create some weird cases when inserting comments in/after properties - for example, node null//=true TODO: make this false maybe? might see null//=true as a property and then fail to parse the following comment. Especially when debugging, I imagine users that don't know about the /- syntax (I'd never seen it before!) could try to comment the rest of the node without a preceding space and run into these problems. (e.g. node// property=0)

Since string identifiers do exist (although I have yet to add them in my implementation!), I'm interested in your thoughts on making them a requirement for using any comment syntax within an identifier, i.e. "iden/*tifier"=0, and simply allowing comments anywhere else in the document.

[zkat]

I think this will be addressed by #75