github github
stars 376
issues 6
subscribers 5
forks 11



11 days ago


Just ask an otter! 🦦

What is otter.nvim?

tldr: Otter.nvim provides lsp features and a code completion source for code embedded in other documents


When implementing autocompletion, code diagnostics and the likes for quarto-nvim I realized that a core feature would be useful to other plugins and usecases as well. quarto documents are computational notebooks for scientific communication based on pandocs markdown. One key feature is that these qmd documents can contain exectuable code blocks, with possibly different languages such as R and python mixed in one document.

How do we get all the cool language features we get for a pure e.g. python file -- like code completion, documentation hover windows, diagnostics -- when the code is just embedded as code blocks in a document? Well, if one document can't give us the answer, we ask an otter (another)! otter.nvim creates and synchronizes hidden buffers containing a single language each and directs requests for completion and lsp requests from the main buffer to those other buffers (otter buffers).

Example in a markdown (or quarto markdown) document

# Some markdown
Hello world

import numpy as np

We create a hidden buffer for a file

import numpy as np

This contains just the python code and blank lines for all other lines (this keeps line numbers the same, which comes straight from the trick that the quarto dev team uses for the vs code extension as well). Language servers can then attach to this hidden buffer. We can do this for all embedded languages found in a document.

A group of otters is called a raft

Each otter-activated buffer can maintain a set of other buffers synchronized to the main buffer.

In other words, each buffer can have a raft of otters!

otter.nvim contains a completion source for nvim-cmp. When a completion request is made, the whole raft of otters is synchronized and the question is directed to the otters. The accumulated answers are then displayed in the main buffer.

    Main --> 🦦1
    Main --> 🦦2
    Main --> 🦦3
    🦦1 --> response
    🦦2 --> response
    🦦3 --> response
    response --> Main: filter

How do I use otter.nvim?

The easiest way to get started is try it out in quarto-nvim or look at the usecases there. Specifically, you'll want to look at the lua/plugins/quarto.lua file in the quarto-nvim-kickstarter configuration.

In short:

Configure otter

If you want to use the default config below you don't need to call setup.

local otter = require'otter'
  lsp = {
    hover = {
      border = { "╭", "─", "╮", "│", "╯", "─", "╰", "│" },
    -- `:h events` that cause the diagnostics to update. Set to:
    -- { "BufWritePost", "InsertLeave", "TextChanged" } for less performant
    -- but more instant diagnostic updates
    diagnostic_update_events = { "BufWritePost" },
  buffers = {
    -- if set to true, the filetype of the otterbuffers will be set.
    -- otherwise only the autocommand of lspconfig that attaches
    -- the language server will be executed without setting the filetype
    set_filetype = false,
    -- write <path>.otter.<embedded language extension> files
    -- to disk on save of main buffer.
    -- usefule for some linters that require actual files
    -- otter files are deleted on quit or main buffer close
    write_to_disk = false,
  strip_wrapping_quote_characters = { "'", '"', "`" },
  -- Otter may not work the way you expect when entire code blocks are indented (eg. in Org files)
  -- When true, otter handles these cases fully. This is a (minor) performance hit
  handle_leading_whitespace = false,

Configure autocompletion

Apart from its own functions, otter.nvim comes with a completion source for nvim-cmp for the embedded code. Use it as follows:

local cmp = require'cmp'
    -- <rest of your nvim-cmp config>
    sources = {
        { name = "otter" },
        -- <other sources>

Activate otter

Activate otter for the current document with

-- table of embedded languages to look for.
-- default = nil, which will activate
-- any embedded languages found
local languages = {'python', 'lua' }

-- enable completion/diagnostics
-- defaults are true
local completion = true
local diagnostics = true
-- treesitter query to look for embedded languages
-- uses injections if nil or not set
local tsquery = nil

otter.activate(languages, completion, diagnostics, tsquery)

Use otter

Assuming otter.nvim is configured and added to nvim-cmp as a completion source, call require'otter'.activate({'python', 'r', <more languages you want to embed> }) on any buffer that has injections (see :h treesitter-language-injections) defined and you will see code completion and diagnostics (on save).

Then use the otter.ask_... functions to e.g. ask for hover documentation, references or the definition.

otter.ask_ functions fall back to regular lsp requests on the main buffer when not in an otter context. Alternatively, you can pass a custom fallback to use instead of the default lsp request. Currently implemented functions are:


Additional functions:

-- Export the raft of otters as files.
-- Asks for filename for each language.


otter.nvim requires the following plugins:

  'hrsh7th/nvim-cmp', -- optional, for completion

An otter eagerly awaiting your lsp requests.