feature: distinguish comment payload from comment markers

[BrunoMSantos]

Did you check the tree-sitter docs?

• I have read all the tree-sitter docs if it relates to using the parser

Is your feature request related to a problem? Please describe.

I'm trying to inject other parsers onto C/C++ comments where multi-line context is crucial for correct highlight, namely rst.

Yet I (1) found no way of creating a capture that takes only the relevant parts of the comment and I suspect it is not possible to do so (not sure though) and (2) it's way harder and prone to difficulties than for (e.g.) Python where one has '(expression_statement (string ...))`.

A common injection for C comments is doxygen which is already meant to parse C/C++ comments, but not rst. Closest I got was something like:

 ((comment) @injection.content                              
   (#lua-match? @injection.content "^/[*][*]<?[^*].*[*]/$") 
   (#offset! @injection.content 0 3 0 -2)                   
   (#set! injection.language "rst"))                        

Which kind of works, but will fail for comments like this, as an example:

/**
 * Some title <- seen as a bullet point due to the leading `*` being part of the capture
 * ==========
 */

Describe the solution you'd like

A capture that sensibly excludes comment leads, ends and prefixes, and yields the 'payload' only. I'm assuming this is possible of course, which maybe it isn't and maybe I actually need a new rst parser just for this.

Describe alternatives you've considered

Somehow recreating this with a lot of regexes, but I have yet to figure out how and, again, I'm not sure this is even possible one way or another without a specific parser from my little knowledge of how tree sitter works.

Additional context

This alternative to Doxygen: https://github.com/jnikula/hawkmoth