[mini.completion] No ctrl-o to escape insert mode when popup is active?

[abeldekat]

Contributing guidelines

• I have read CODE_OF_CONDUCT.md

Module(s)

mini.completion

Question

no_ctrl_o.mp4

init.lua

---@diagnostic disable: undefined-global
local function setup_deps(path_to_site)
  local mini_path = path_to_site .. "pack/deps/start/mini.nvim"
  if not vim.uv.fs_stat(mini_path) then
    vim.cmd('echo "Installing `mini.nvim`" | redraw')
    vim.fn.system({ "git", "clone", "--filter=blob:none", "https://github.com/echasnovski/mini.nvim", mini_path })
    vim.cmd("packadd mini.nvim | helptags ALL")
    vim.cmd('echo "Installed `mini.nvim`" | redraw')
  end
  require("mini.deps").setup({ path = { package = path_to_site } })
end

local function setup_native_autocompletion()
  vim.api.nvim_create_autocmd("LspAttach", {
    group = vim.api.nvim_create_augroup("my.lsp", {}),
    callback = function(args)
      local client = assert(vim.lsp.get_client_by_id(args.data.client_id))
      if client:supports_method("textDocument/completion") then
        local chars = {} -- optional: trigger autocompletion on EVERY keypress. May be slow!
        -- stylua: ignore
        for i = 32, 126 do table.insert(chars, string.char(i)) end
        client.server_capabilities.completionProvider.triggerCharacters = chars
        vim.lsp.completion.enable(true, client.id, args.buf, { autotrigger = true })
      end
    end,
  })
end

local function setup_mini_autocompletion()
  require("mini.snippets").setup() -- only using lsp snippets...
  require("mini.completion").setup({ lsp_completion = { auto_setup = false, source_func = "omnifunc" } })
end

local function custom_on_attach(_, bufnr) vim.bo[bufnr].omnifunc = "v:lua.MiniCompletion.completefunc_lsp" end

local use_mini = false
vim.g.mapleader = " "
vim.o.shortmess = "CFOSWaco"
vim.o.completeopt = "menuone,noselect,fuzzy"

setup_deps(vim.fn.stdpath("data") .. "/site/")
MiniDeps.now(function()
  require("mini.icons").setup()
  MiniIcons.tweak_lsp_kind()
  MiniDeps.add({ source = "nvim-treesitter/nvim-treesitter", checkout = "master" })
  require("nvim-treesitter.configs").setup({ ensure_installed = { "rust" } })

  if use_mini then
    setup_mini_autocompletion()
  else
    setup_native_autocompletion()
  end

  MiniDeps.add("neovim/nvim-lspconfig")
  require("lspconfig").rust_analyzer.setup({ -- have rust-analyzer installed...
    on_attach = use_mini and custom_on_attach or nil,
  })
end)

rust code

// start a simple project with "cargo new demo"
// cd demo, and edit src/main.rs
fn main() {
    println!("Hello, world!");
}

struct Dog<Breed> {
    name: String,
}

I like working with the code snippets provided by the lsp. However, sometimes they can be hard to use.

In the video, the lsp returns the following snippet:

{ body = "PhantomData<$0>" }

Inside the $0, a generic can be placed, T for example. After doing so, I have to "overcome" the closing > in order to be able to continue typing. See the video, at "00:15" and "00:40".

When using native autocompletion, I can type <c-o>a and continue. Alternatively, I am also able to avoid snippet activation by not pressing <c-y>.

With mini, when I type <c-o>, the selection in the popup moves one down instead of changing the mode. There is no way to have the word "PhantomData" completed without also activating the snippet(by design...).

Would it be possible to have <c-o> behave as in native autocompletion?

Sidenote: A much better lsp snippet would have been:

{ body = "PhantomData<$1>$0" } 

[echasnovski]

Would it be possible to have <c-o> behave as in native autocompletion?

I am afraid the answer to the particular question is "no, if you want to keep using 'omnifunc'". This is because <C-o> key is very hard-coded after <C-x><C-o> is pressed (i.e. 'omnifunc' completion is triggered, which is what 'mini.completion' is doing). As in "it is currently impossible to override in this situation, at all". It is documented in :h i_CTRL-X_CTRL-O and mentioned in the help (although in a slightly different place).

The reason this doesn't happen with built-in completion is because it uses vim.fn.complete() directly. The downside of this is that it won't re-trigger on <BS> and implementing it requires non-trivial workarounds (which built-in completion still hasn't done yet). The re-trigger on <BS> works with 'completefunc' and 'omnifunc', which (if I remember correctly) was the main reason to even choose this 'completefunc'/'omnifunc' in the first place about 4 years ago.

The are several possible solutions to the problem:

• Use lsp_completion.source_func = 'completefunc' (with or without auto_setup), but this will "disable" the <C-u> key (for the same reason, see :h i_CTRL-X_CTRL-U) which I find more critical.
• Use mappings similar to move_with_alt from 'mini.basics' to move in Insert/Command-line mode. I personally prefer this solution.
• Create an issue in 'rust-analyzer' to have snippets follow "PhantomData<$1>$0" design.
• Somehow manually modify such snippets in lsp_completion.process_items. This might get complicated, but should be doable.

[abeldekat]

Thank you for taking the time to answer my question so thoroughly!
For now, I will use completefunc, as I never use I_CTRL-U.

I do have the mini.basics mappings. Automating their usage will require some training...)

I am linking an earlier question as well as the answer is a bit related:

[mini.completion] Completefunc or omnifunc?

[echasnovski]

I do have the mini.basics mappings. Automating their usage will require some training...)

For me at least 90% of their usage is with <M-l> to move once out of some paired character. Usually even if 'mini.pairs' supports it (like ), etc.), just because a single mapping leads to less mental overhead.