New release and roadmap

[jalvesaq]

R.nvim is quite stable lately. I think it's time for a new release...

Does anyone have plans for new features?

[jalvesaq]

Suggested changelog for the new release (including only the most relevant changes visible to users):

- Restore the debugging feature.

- Better integration with TreeSitter:
  - Better code chunk extraction from RMarkdown or Quarto files.
  - Enhance piped commands handling in R code.
  - Refactor the number formatting logic.
  - Optimize subsetting.

- New options: `yaml_hl`, `set_params`.

- When editing a Markdown document, behave as if it were an RMarkdown file if
  "markdown" is included in the list `vim.g.R_filetypes`.

- Support for Wezterm and Kitty split panes.

- Create more functions that run hooks:
  - `run_after_R_start`.
  - `run_after_ob_open`.

- Add browsing and completion of S7 objects (#429)
    
- Add lua/r/health.lua (`:checkhealth r`)

- Replace some options with default commands.
  - Remove assignment_keymap, pipe_keymap, rmd_chunk_keymap and rnw_chunk_keymap.
  - Rename some `<Plug>` labels:
      - RAssign as RInsertAssign
      - RPipe as RInsertPipe
      - RmdCreateChunk as RmdInsertChunk
      - RnwCreateChunk as RnwInsertChunk

- Remove option `is_darwin`.

- Remove support for rgui.exe.
    
- Remove integration with Rapp on macOS.
    
- Bug fixes.

[gueyenono]

This comment does not belong in this thread, but I just wanted to say thanks to you and all contributors for such a great plugin.

[PMassicotte]

I think this is a good moment for a new release. I have not realized, but like you mentioned, many new things have been added the last few months and no issues appeared.

[PMassicotte]

Looking at the log:

• Health check improvements: Fixed treesitter parsing issues and parser detection in health checks (fix(health.lua): add delay to ensure treesitter parsing completion before proceeding #434, fix(health.lua): ensure parser_installed checks for has_parser function availability #433)
• Documentation command fix: Resolved "command not found" error with LibreOffice Writer in documentation (fix word document not opening in libreoffice after rendering markdown #431)
• Label attribute validation: Added proper checking for character vector label attributes (Check if label attribute is a character vector #427)
• Fuzzy search fix: Resolved overflow issues in fuzzy completion (Fix fuzzy overflow #416)
• Roxygenize function fix: Corrected issues with the roxygenize function (fix roxygenize function #412)
• Paste lines limit: Fixed handling of max_paste_lines when sending functions (Consider max_paste_lines when sending function #409)

[jalvesaq]

So, the changelog becomes:

- Restore the debugging feature.

- Better integration with TreeSitter:
  - Better code chunk extraction from RMarkdown or Quarto files.
  - Enhance piped commands handling in R code.
  - Refactor the number formatting logic.
  - Optimize subsetting.

- New options: `yaml_hl`, `set_params`.

- When editing a Markdown document, behave as if it were an RMarkdown file if
  "markdown" is included in the list `vim.g.R_filetypes`.

- Support for Wezterm and Kitty split panes.

- Create more functions that run hooks:
  - `run_after_R_start`.
  - `run_after_ob_open`.

- Add browsing and completion of S7 objects (#429)
    
- Add lua/r/health.lua (`:checkhealth r`)

- Replace some options with default commands.
  - Remove assignment_keymap, pipe_keymap, rmd_chunk_keymap and rnw_chunk_keymap.
  - Rename some `<Plug>` labels:
      - RAssign as RInsertAssign
      - RPipe as RInsertPipe
      - RmdCreateChunk as RmdInsertChunk
      - RnwCreateChunk as RnwInsertChunk

- Remove option `is_darwin`.

- Remove support for rgui.exe.
    
- Remove integration with Rapp on macOS.
    
- Bug fixes:
  - Health check improvements: Fixed treesitter parsing issues and parser
    detection in health checks (#434, #433).
  - Documentation command fix: Resolved "command not found" error with
    LibreOffice Writer in documentation (#431).
  - Label attribute validation: Added proper checking for character vector
    label attributes (#427).
  - Fuzzy search fix: Resolved overflow issues in fuzzy completion (#416).
  - Roxygenize function fix: Corrected issues with the roxygenize function (#412).
  - Paste lines limit: Fixed handling of max_paste_lines when sending functions (#409).
  - Other minor bug fixes.

[jalvesaq]

We have to define the version number, which depends on the roadmap. We have some known bugs listed in the documentation and some open issues that we could not fix. What are the most important missing features and existing bugs that prevent the plugin from deserving the 1.0 version number? Does anyone plan to implement new features or fix existing known bugs? Regardless of plans for new features and bug fixes, should we wait for Neovim 1.0 to release our 1.0 version?

[PMassicotte]

I think we covered all the low-hanging fruit at this time. Do you see anything we could try to tackle?

[jalvesaq]

We don't have completion for the YAML fields of a Quarto document metadata, but this should be possible with yamlls and the use of a separate _quarto.yml file. I added this modeline to a _quarto.yml but it didn't work:

# yaml-language-server: $schema=/opt/quarto/share/editor/tools/yaml/yaml-intelligence-resources.json

Other than that, we lack a good debugger. But, after 16 years since the first release of Vim-R-plugin, I'm a bit tired and have nothing to offer as roadmap. I suggest 0.99 as the version number and, after a few months, version 1.0.

[PMassicotte]

I have a semi working version for completing yml. Opening an issue.

[jalvesaq]

Good job! Let's wait for this before a new release!
I recently renamed Nvim-R as Vim-R. I'll do a pull request updating the references to Vim-R in R.nvim too.

[jalvesaq]

Today, I looked again at R's C code, but I couldn't find a way of updating the Object Browser with the objects of a function being debugged. As far as I could see, it would be necessary to write a front-end for R (as RStudio does). I don't plan to do this and, unless, someone wants to try to develop such a front-end, our current debugging feature has no path for a substantial improvement.

I'm not badly missing any other feature. So, I propose the following simple roadmap:

• Version 0.99.0 after merging the YAML completion feature.

• Versions 0.99.x for some time:

  • No planed new features for these versions, only bug fixes, code optimization, and improvements in the documentation and the Wiki.

  • The Wiki should show an example of how to build a complete IDE for R using tree-sitter (code highlighting, indentation and folding), LSP (diagnostics, completion), terminalgraphics (plot in the R Console), and R.nvim (completion, object browser, sending code to R, etc).

  • Require Neovim 0.11.5 or even 0.12.0 as the minimal version, and delete code with deprecated functions.

  • If vim.lsp.server() is implemented, evaluate the possibility of using it instead of nvim-cmp for code completion.

• Version 1.0.0 after the release of Neovim 0.11.5 or 0.12.0.

[PMassicotte]

That makes a lot of sense. We should do that.

[jalvesaq]

I'm learning how to make an internal LSP: https://github.com/jalvesaq/zotcite/blob/lsp/lua/zotcite/lsp.lua

I'll start to develop a similar feature in R.nvim soon. Then, it will work with either blink.cmp or nvim-cmp (and cmp-r will become obsolete).

[PMassicotte]

That is pretty nice. Let me know, very looking to learn how to make an lsp.

[jalvesaq]

I added an "lsp" branch. Completion of R objects is working with both nvim-cmp and blink.nvim, but the extra details (documentation) only appear in nvim-cmp. I can't waste time implementing the other types of completion if this bug isn't solved...

[PMassicotte]

I am not sure to understand. Some completions are not working in blink?

[PMassicotte]

Do you mean labels that are shown on the right side of the completion items? I can try it tomorrow.

[jalvesaq]

Now that the pull request #440 is merged, I'll release R.nvim 0.99.0...

[jalvesaq]

I released version 0.99.0, but, looking at 0.1.0 release notes, I see a missing section: "New contributors". I'll edit the release notes to add the missing section...

[jalvesaq]

The completion works. But when an item in the menu is selected, additional information is displayed in the pop-up window. This pop-up window doesn't appear when Blink is the completion plugin.

[PMassicotte]

Are we supposed to get function args completion?

[samtrek]

I have a similar issue with zotcite, also have the issue with shortening of the home to ...e, for the file path, no matter what I do, I could not get to work on my laptop, but works fine on the desktop, could it be due to the different file system, for my laptop I am using btrfs and the destkop ext4, nothing comes to my mind other than this.
I have gotten R. nvim to work now by manually pointing to the path. Seems like a patch up solution for now.

[PMassicotte]

One thing I just noticed, I can <leader>rq but not <leader>rf afterward.

It's fixed now.

Confirmed, it works. Tx

[jalvesaq]

could it be due to the different file system

It turns out that short_src from debug.getinfo() is limited to 60 characters. Does it work now (using source)?

[samtrek]

yes, now works has expected, could you look into the zotcite plugin also, having the same issue, thank you.

[jalvesaq]

Yes, thanks!

[samtrek]

1. The new lsp completion does not show completions when a second r file is opened in a new buffer.
2. When a previously opened file is is restored the completion also does not work (lazyvim restore function default settings), nvimcomr does not start and cannot also start nvimcomr manually.
   These are my observations for now.
   Config
   using an external kitty term with radian as the r terminal

[PMassicotte]

Thank you for your feedback!

[PMassicotte]

I also have no completion when multiple buffers are opened.

[jalvesaq]

Thanks, @samtrek and @PMassicotte !

It should attach to all buffers now. I no longer use lazy.nvim to replicate the restore bug, but I may look at this later.

[samtrek]

I dont know if it is my setup, unless sent to the terminal their is no completion from already written r script or codes

to replicate

1. Open any previous r script
2. use a function or any other variables from the script without sending to term
3. Only r default functions and possible variables are given as completion and none from the script

I hope it can be simulated

expected
the already written script should be also be used as a completion by nvimcomr (my opinion and the previous default)

[jalvesaq]

Before starting R, R.nvim only looks for library commands to determine which functions to make available for completion. It doesn't do any other code analysis. It inspects R's memory. You can see this as a feature or a limitation. Please, see #95

You can disable R.nvim's completion, signature help, and hover, and use these features from the languageserver. To disable the features from R.nvim:

require("r").setup({
    r_ls = {
        completion = false,
        hover = false,
        signature = false,
    }
})

[samtrek]

Then as a feature then, no I prefer not to use languageserver, prefer to use the treesitter