quarto, rmd no longer registered in nvim-treesitter

[PMassicotte]

nvim-treesitter/nvim-treesitter@da61d31

One way to revert it back:

vim.treesitter.language.register("markdown", { "quarto", "rmd" })

[jmbuhr]

Oh, interesting. I assume this means that associations of this kind should live in their respective plugins, but we can check back with @clason on the commit about the intended direction of the change. I can't do so right now, because commenting on commits is not loading for me on mobile. Would you do that?

I can add the fix to quarto-nvim. Did you do so for Rmd via R.nvim?

Along the same vein we should look towards creating a treesitter grammar for Rmd and Qmd to accommodate the differences from plain markdown. There is a way to do inheritance for grammars, but I didn't have the time to look into it more deeply.

[clason]

This is intentional since this breaks the automated commentstring detection in Neovim.

[jmbuhr]

Ok, thanks! Will the commentstrings also break if we add the association via vim.treesitter.language.register or is this indeed the intended way?

[clason]

I believe it would also break, at least on Neovim master.

Rmd/quarto is not Markdown, even though it looks very similar to the untrained eye. You really should write a dedicated parser for it. (It's possible to extend other parsers so you don't have to start from scratch.)

[PMassicotte]

I can add the fix to quarto-nvim. Did you do so for Rmd via R.nvim?

I have initiated a PR here: R-nvim/R.nvim#239

It seems to be working fine with the little testing I have done so far.

[PMassicotte]

That is also interesting that I am using qmd and md a lot in my day-to-day work and never had this commentstrings issue. I am not sure what in my config prevented this to happen.

[jmbuhr]

Agreed, I started with a Qmd parser at some point, just as a fork of the markdown one, but didn't have the time to finish it properly.

[jmbuhr]

Relying on the markdown grammar was always just a temporary solution that happend to hold for a surprising amount of time, so I was anticipating this.

[PMassicotte]

I agree. We have patched this in R.nvim, but we should properly extend markdown tree-sitter.

[jmbuhr]

I suppose we can call it pandoc-markdown, as Rmd and Qmd are both just pandoc's markdown with most native pandoc extensions enabled plus some custom ones that only add and don't conflict with base pandoc's markdown.

https://pandoc.org/MANUAL.html#pandocs-markdown

[clason]

If the difference is only the comment string (or at least only at the top level), you could just replace that and still keep the markdown_inline injections. That would vastly simplify things.

[jmbuhr]

In the short term this will be my first test once I'm back from vacation, though in the long run we'll want to extend both the block and inline parser to accommodate differences (e.g. surrounding ~ denote subscript and ~~ denote strikethrough in pandoc's markdown, while the commonmark grammar recognizes both as strikethrough).

[clason]

Well, there was an "if" in this statement ;)

[PMassicotte]

@jmbuhr Let me know, I would be interested contributing.

[jmbuhr]

@PMassicotte Amazing! I'll add you once I have a repo.