vim-gutentags VS nvim-cmp

https://github.com/ludovicchabant/vim-gutentags/issues/139 has some background.

Hi everyone. I'm looking to have a better setup for Rust in neovim. I do have rust-analyzer installed for useful lsp things but I was hoping to get tags working for it as well. I was using vim-gutentags (https://github.com/ludovicchabant/vim-gutentags) for navigating useful functions and stuff but couldn't quite get it to work for rust. Is there a simple way to do it or do I need rusty-tags and some aucommand to get it to work?

use tags, I like https://github.com/ludovicchabant/vim-gutentags for this. I might use my local .vimrc to tweak the config (exclude compiled source files and other uninteresting things)

guttentag: https://github.com/ludovicchabant/vim-gutentags

Without more information, it's hard to point you in the right direction. The tags file could be out-of-date, in which case you can try to re-generate it (vim-gutentags for tags auto-generation). You could have 2 function declarations with the same name, in which case you can try :tag to cycle through tags (supports partials, like :tag F which will suggest FOO, FAR, FAB ...etc) or :tag to see a list of possible options (supports partials, like :tag F which will list FOO, FAR, FAB, ...etc) for various matching tags you can jump to (fzf.vim provides a tags fuzzy finder via :Tags). Maybe you're experiencing :h tag-priority?

ludovicchabant/vim-gutentags - Tags

You can then use :cnext and :cprev (or focusing the window and selecting an entry) to navigate between them.

As others have stated, you can also use ctags (plugins like https://github.com/ludovicchabant/vim-gutentags are useful for refreshing tags in a project), but for some languages you may need to add a tag definition (e.g. for something like rust or zig). For older languages like C you should be fine.

Not sure if it will help in this case, but I also depend on ctags for when the lsp fails (e.g. code it doesn't compile for some reason). Here you let a program create a tags file, e.g. I use https://github.com/universal-ctags/ctags (it seems it can parse json files, though I'm not sure what kind of tags are generated from this and if they will be useful to you) with https://github.com/ludovicchabant/vim-gutentags to update the tags file. The tags file just contains symbol names with locations where they are defined, and vim has builtin functionality to use these tags files :h tag and they (can) work filetype independent. For example if I mention a C type in a markdown document I can just use ctrl-[ to jump to its definition in the C source file. Possibly you can generate tags files yourself from the json files to help with this, the tags file format is not very complicated. Tags also are not very intelligent and depend on unique names for them to work well, there is the :h g_CTRL-] that can help, but for symbols that are very common (e.g. init or something that potentially has like 20+ definitions) it doesn't really work.

vim.api.nvim_set_hl(0, "PmenuSel", { bg = c.background_light, fg = "NONE" }) vim.api.nvim_set_hl(0, "Pmenu", { fg = c.foreground, bg = c.background_light }) vim.api.nvim_set_hl(0, "CmpItemAbbrDeprecated", { fg = c.foreground_light, bg = "NONE", strikethrough = true }) vim.api.nvim_set_hl(0, "CmpItemAbbrMatch", { fg = c.blue , bg = "NONE", bold = true }) vim.api.nvim_set_hl(0, "CmpItemAbbrMatchFuzzy", { fg = c.blue, bg = "NONE", bold = true }) vim.api.nvim_set_hl(0, "CmpItemMenu", { fg = c.purple, bg = "NONE", italic = true }) vim.api.nvim_set_hl(0, "CmpItemKindField", { fg = c.red }) vim.api.nvim_set_hl(0, "CmpItemKindProperty", { fg = c.red }) vim.api.nvim_set_hl(0, "CmpItemKindEvent", { fg = c.red }) vim.api.nvim_set_hl(0, "CmpItemKindText", { fg = c.green }) vim.api.nvim_set_hl(0, "CmpItemKindEnum", { fg = c.green }) vim.api.nvim_set_hl(0, "CmpItemKindKeyword", { fg = c.green }) vim.api.nvim_set_hl(0, "CmpItemKindConstant", { fg = c.yellow }) vim.api.nvim_set_hl(0, "CmpItemKindConstructor", { fg = c.yellow }) vim.api.nvim_set_hl(0, "CmpItemKindReference", { fg = c.yellow }) vim.api.nvim_set_hl(0, "CmpItemKindFunction", { fg = c.purple }) vim.api.nvim_set_hl(0, "CmpItemKindStruct", { fg = c.purple }) vim.api.nvim_set_hl(0, "CmpItemKindClass", { fg = c.purple }) vim.api.nvim_set_hl(0, "CmpItemKindModule", { fg = c.purple }) vim.api.nvim_set_hl(0, "CmpItemKindOperator", { fg = c.purple }) vim.api.nvim_set_hl(0, "CmpItemKindVariable", { fg = c.foreground }) vim.api.nvim_set_hl(0, "CmpItemKindFile", { fg = c.foreground }) vim.api.nvim_set_hl(0, "CmpItemKindUnit", { fg = c.orange }) vim.api.nvim_set_hl(0, "CmpItemKindSnippet", { fg = c.orange }) vim.api.nvim_set_hl(0, "CmpItemKindFolder", { fg = c.orange }) vim.api.nvim_set_hl(0, "CmpItemKindMethod", { fg = c.blue }) vim.api.nvim_set_hl(0, "CmpItemKindValue", { fg = c.blue }) vim.api.nvim_set_hl(0, "CmpItemKindEnumMember", { fg = c.blue }) vim.api.nvim_set_hl(0, "CmpItemKindInterface", { fg = c.aqua }) vim.api.nvim_set_hl(0, "CmpItemKindColor", { fg = c.aqua }) vim.api.nvim_set_hl(0, "CmpItemKindTypeParameter", { fg = c.aqua }) -- https://github.com/hrsh7th/nvim-cmp/pull/1689 vim.api.nvim_set_hl(0, 'FloatBorder', { fg = c.background_light, bg = c.background_light })

Doing a fast scan on the commit history this is probably related to https://github.com/hrsh7th/nvim-cmp/pull/1689 . Hope it helps.

https://github.com/hrsh7th/nvim-cmp This is an autocompletion engine https://github.com/nvim-treesitter/nvim-treesitter This allows NeoVim to install parsing scripts so NeoVim can do things like code highlighting. https://github.com/williamboman/mason.nvim Not strictly necessary, but allows you to access a repo of LSP, install them, and configure them for without you actively messing about in config files. https://github.com/neovim/nvim-lspconfig Also not strictly necessary, but vastly simplifies LSP setup. https://github.com/williamboman/mason-lspconfig.nvim This lets the above two plugins talk to each other more easily.

Edit: Solved I only had to do this

toggling sources by Treesitter context

But today I'm a different person, and I'm sure that this is the right decision. I already gave a spiel on the cmp PR, but lemme copy-paste my response for those that weren't following the issue or don't check their GitHub notifications:

EDIT: I was able to reproduce the issue with a simplified nvim-cmp init.vim - posted here: https://github.com/hrsh7th/nvim-cmp/issues/1728

You need to add nvim-cmp and cmp-nvim-lsp

one thing that always annoyed me is how the autocomplete window behaves. when the space below your current line isn't sufficient it just teleports above the line. i know you can set a custom menu direction but i'd much rather have it pop up in a consistent place that is below the current line. the other behavior i dislike is the sizing. it's as big as the longest function in the list which can make the window enormous, to the point where there's almost no place left for the docs window. there aren't really any docs in the example below but i imagine with the little space that is left it'll get troublesome for more verbose docs such as with rust. i'd like to somehow tell nvim-cmp to always pop up below the current line, no matter how long the list is and possibly limit the horizontal size. is that at all possible? i've looked through the nvim-cmp wiki but haven't found an adequate solution. does this go beyond nvim-cmp?

Update: This isn't currently supported by nvim-cmp (and in fact it has been a feature request for a bit) and so I did my best and opened a PR that implements it :)