Is it possible to support comment.inner?

[Freed-Wu]

https://github.com/glts/vim-textobj-comment support inner comment and outer comment.

For this project, many textobjects.scm only have comment.outer. Is it possible to add support for comment.inner?

[sharpchen]

@comment.inner is already one of the builtin conventions

nvim-treesitter-textobjects/BUILTIN_TEXTOBJECTS.md

Line 13 in 43a2b80

13. @comment.inner

[Freed-Wu]

I mean many textobjects.scm only have comment.outer.

[sharpchen]

Yeah it might be impossible for those languages as the parser doesn't produce a node for the inner content.

[Freed-Wu]

Can we have a guide about how to name the node? such as:

grammar.js:

    comment: $ => seq('#', alias(/[^\n]*/, $.body)),

queries/highlights.scm:

(comment) @comment

(comment
  (body) @spell)

queries/injections.scm:

(comment
  (body) @injection.content
  (#set! injection.language "comment"))

queries/textobjects.scm:

(comment
  (body) @comment.inner) @comment.outer

Then there is a guide for all users of grammars to respect it.

[sharpchen]

Do you mean a conventional (type)name for comment content node on parser implementation?
That won't bring much benefit, because you have to manage query in dedicated file for each language anyway.
The problem is the parser doesn't generate a inner node for comment, not how we write queries.

[Freed-Wu]

If we need to realize comment.inner, should it contains space?

    comment: $ => seq('#', alias(/[^\n]*/, $.body)),

or

    comment: $ => seq(/#\s*/, alias(/[^\n]*/, $.body)),

So a guideline to tell the writer of grammars is necessary.

[sharpchen]

The problem is the parser doesn't generate a inner node for comment, not how we write queries.

I should take this back, it's actually possible to set capture without extra nodes, by using treesitter predicates and directives. See

nvim-treesitter-textobjects/queries/python/textobjects.scm

Lines 32 to 35 in 7359dfc

	; leave space after comment marker if there is one
	((comment) @comment.inner @comment.outer
	(#offset! @comment.inner 0 2 0 0)
	(#lua-match? @comment.outer "# .*"))

, :h treesitter-predicates and :h treesitter-directives

If we need to realize comment.inner, should it contains space?

Given the availability of treesitter predicates and directives, the parse can either contain space or not.
It should be kept if the doc comment contains certain format by indentation, otherwise when user will get inner content with broken format when copying from the range.

[Freed-Wu]

it's actually possible to set capture without extra nodes

Cool. So all grammars support it. We should add comment.inner for more grammars.

[sharpchen]

We should add comment.inner for more grammars.

Yes and no, if the language has both blockwise comment and linewise comment, but without comment type distinction, we can't really set @comment.inner for it even with predicates and directives.

The expressiveness of treesitter predicates and directives are limited, so we can't use them to differentiate the type depending entirely on the (outer)node content, if the language has more than one comment type but not distinction for them

We can do that for python because there's no blockwise comment in the language.

The current queries supporting @comment.inner, its target parser either has comment type distinction(like line_comment and block_comment) or has only one comment type(like python and fish), or with extra child node for the inner content(like lua)