nvim-cmp stops working when having vim-dadbod-completion installed about vim-dadbod-completion HOT 10 CLOSED

I'm also using Lazy. You can try my setup. Here's how I set up dadbod related stuff: https://github.com/kristijanhusak/neovim-config/blob/master/nvim/lua/partials/plugins/db.lua

And this is where I set up cmp for it: https://github.com/kristijanhusak/neovim-config/blob/master/nvim/lua/partials/plugins/completion.lua#L93-L106

from vim-dadbod-completion.

Thanks! It was not having the "cmd" line in my case. Apart from changing vsnip to luasnip copy pasting the second worked like a charm as well.

Reference solution for anyone else using LazyVim that finds this issue.

Thanks for the help and the great applications! @kristijanhusak
Keep up the great work!

-- plugins/cmp.lua
return {
  {
    "hrsh7th/nvim-cmp",
    lazy = true,
    opts = function(_, opts)
      local cmp = require("cmp")
      opts.completion = {
        completeopt = "menu,menuone,noinsert",
      }
      opts.window = {
        completion = cmp.config.window.bordered(),
        documentation = cmp.config.window.bordered(),
      }
      return opts
    end,
  },
}

-- plugins/dadbod.lua
return {
  {
    "kristijanhusak/vim-dadbod-ui",
    dependencies = {
      { "tpope/vim-dadbod", lazy = true },
      { "kristijanhusak/vim-dadbod-completion", lazy = true, ft = { "sql", "mysql", "plsql" } },
    },
    cmd = { "DBUI", "DBUIFindBuffer" },
  },
}

-- config/autocmds.lua
local autocomplete_group = vim.api.nvim_create_augroup("vimrc_autocompletion", { clear = true })
local cmp = require("cmp")
vim.api.nvim_create_autocmd("FileType", {
  pattern = { "sql", "mysql", "plsql" },
  callback = function()
    cmp.setup.buffer({
      sources = {
        { name = "vim-dadbod-completion" },
        { name = "buffer" },
        { name = "luasnip" },
      },
    })
  end,
  group = autocomplete_group,
})

from vim-dadbod-completion.

I have also had this issue.

return {
  {
    "tpope/vim-dadbod",
    dependencies = {
      "kristijanhusak/vim-dadbod-ui",
    },
  },
  {
    "kristijanhusak/vim-dadbod-ui",
  },
  -- {
  --   "kristijanhusak/vim-dadbod-completion",
  -- },
}

Commenting and uncommenting this line toggles the issue.

I am using LazyVim and my nvim-cmp config is completely default apart from a slight style choice.

return {
  {
    "hrsh7th/nvim-cmp",
    lazy = true,
    opts = function(_, opts)
      local cmp = require("cmp")
      opts.completion = {
        completeopt = "menu,menuone,noinsert",
      }
      opts.window = {
        completion = cmp.config.window.bordered(),
        documentation = cmp.config.window.bordered(),
      }
      return opts
    end,
  },
}

I would love this behavior inside of SQL editors but this just disables it globally for me.

I notice a strange behavior where it seems that nvim-cmp loading gets placed "behind" dadbod-completion when it is active.

-- `:Lazy profile` with dadbod-completion
      ➜  after 5.14ms
        ★   after/plugin/vim_dadbod_completion.lua 5.06ms
          ‒   nvim-cmp 4.53ms
            ●   crates.nvim 2.32ms
            ●   copilot-cmp 0.53ms
              ➜   copilot.lua 0.46ms
                ★   copilot.lua/plugin/copilot.lua 0.03ms
            ●   cmp-nvim-lsp 0.04ms
            ●   cmp-buffer 0.08ms
            ●   cmp-path 0.06ms
            ●   cmp_luasnip 0.07ms
            ●   nvim-cmp/plugin/cmp.lua 0.26ms
        ★   after/plugin/vim_dadbod_completion.vim 0.02ms

and removed from "InsertEnter" where I normally see it.
I'm not sure if the extension is taking ownership and overwriting the "event" somewhere. But something along those lines is what the behavior seems to suggest.

-- `:Lazy profile` without dadbod-completion
InsertEnter 21.42ms
      ➜   nvim-cmp 21.31ms
        ★   crates.nvim 10.62ms
          ‒   after/plugin/cmp_crates.lua 0.69ms
        ★   copilot-cmp 3ms
          ‒   copilot.lua 2.67ms
            ●   copilot.lua/plugin/copilot.lua 0.18ms
        ★   cmp-buffer 5.08ms
          ‒   after/plugin/cmp_buffer.lua 4.95ms
        ★   cmp-path 0.57ms
          ‒   after/plugin/cmp_path.lua 0.47ms
        ★   cmp_luasnip 0.66ms
          ‒   after/plugin/cmp_luasnip.lua 0.6ms
        ★   nvim-cmp/plugin/cmp.lua 0.27ms

from vim-dadbod-completion.

I noticed that my dependencies might have caused some confusion. dadbod requiring ui instead of the other way around.

I will invert these tomorrow and see if that fixes the issue on my end.

I tried reviewing all of the places where this extension interacts with cmp and nothing stood out as being a potential cause.

I will let you know if this corrects the behavior or if I find anything else.

from vim-dadbod-completion.

@kristijanhusak Thank you very much for the help. For me the solution was a little weird. I kept my code just the same, but the only thing I added was local cmp = require("cmp") (could be before or after the autocommand). That is, my code is this at the moment:

vim.api.nvim_create_autocmd("FileType", {
  pattern = { "sql", "mysql", "plsql" },
  command = ":lua require('cmp').setup.buffer({ sources = {{ name = 'vim-dadbod-completion' }} })",
})

local cmp = require("cmp")

I don't understand why requiring cmp as a local config variable solves anything. Doing this alone for example won't work:

vim.api.nvim_create_autocmd("FileType", {
  pattern = { "sql", "mysql", "plsql" },
  callback = function()
    require("cmp").setup.buffer({
      sources = {
        { name = "vim-dadbod-completion" },
        { name = "buffer" },
        { name = "luasnip" },
      },
    })
  end,
})

To make it work I MUST require cmp at some point in the code, for example:

vim.api.nvim_create_autocmd("FileType", {
  pattern = { "sql", "mysql", "plsql" },
  callback = function()
    require("cmp").setup.buffer({
      sources = {
        { name = "vim-dadbod-completion" },
        { name = "buffer" },
        { name = "luasnip" },
      },
    })
  end,
})

require("cmp") -- must have this

Maybe it is because LunarVim is doing something funny with cmp? I don't feel it is because of that, but at least this fixed my issue. Maybe doing require("cmp") initializes some stuff? But then why wont the require within the callback do the same?

Either way, thanks for the great plugins @kristijanhusak. They work really well.

from vim-dadbod-completion.

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.

from vim-dadbod-completion.

I'm closing this since it seems that everyone managed to solve the configuration issue.

from vim-dadbod-completion.