Shougo

I think `mini.completion` does not use fuzzy match. If the items are very big from LS, nvim-cmp fuzzy match is very slow.

I have tested it. Hm. It seems slow. Its slowness is from: * The auto completion keyword length is 0(The items are very huge) * If items are huge, the...

You can change `keyword_length` and disable fuzzy match. https://github.com/hrsh7th/nvim-cmp/blob/main/doc/cmp.txt#L493 https://github.com/hrsh7th/nvim-cmp/blob/main/doc/cmp.txt#L539

It is better for me. ```vim " Setup global configuration. More on configuration below. lua

```lua source.complete = function(self, params, callback) local lsp_params = vim.lsp.util.make_position_params(0, self.client.offset_encoding) lsp_params.context = {} lsp_params.context.triggerKind = params.completion_context.triggerKind lsp_params.context.triggerCharacter = params.completion_context.triggerCharacter self:_request('textDocument/completion', lsp_params, function(_, response) vim.print(response) callback(response) end) end ``` I...

Hm... I have created the minimal vimrc. ```vim set rtp+=~/src/nvim-cmp set rtp+=~/src/cmp-buffer set rtp+=~/src/cmp-cmdline "set rtp+=~/src/nvim-lspconfig "set rtp+=~/src/cmp-nvim-lsp set rtp+=~/src/vim-vsnip set rtp+=~/src/cmp-vsnip set cmdheight=0 lua

It seems `set laststatus=0` is needed.

OK. I have get the reason. https://github.com/hrsh7th/nvim-cmp/blob/630cdf7d547c4461ef6d7362c3794a08abfad4fb/lua/cmp/utils/feedkeys.lua#L36 The feedkeys execution causes the problem. It executes `hcall v:lua.cmp.utils.feedkeys.call.run(1)^M` text.

```diff diff --git a/lua/cmp/init.lua b/lua/cmp/init.lua index 4aaf2fe..37687da 100644 --- a/lua/cmp/init.lua +++ b/lua/cmp/init.lua @@ -330,7 +330,7 @@ local on_insert_enter = function() end end autocmd.subscribe({ 'CmdlineEnter' }, async.debounce_next_tick(on_insert_enter)) -autocmd.subscribe({ 'InsertEnter' }, async.debounce_next_tick_by_keymap(on_insert_enter))...

And it is the dup of https://github.com/hrsh7th/nvim-cmp/issues/1382.