VCaml

VCaml is a library for building Neovim plugins in OCaml. Here's an overview of the directory structure:

vcaml
|-- src     -- The `vcaml` library: typed primitives for interacting with Neovim.
|-- plugin  -- The `vcaml.plugin` library: high-level wrappers for common use-cases.
|              For usage examples, see the "example" and "templates" subdirectories.
|-- test    -- Tests of VCaml. The `vcaml.test.helpers` library has functions that make
|              testing VCaml plugins easy.
`-- debug   -- The `vcaml.debug` library: meant to be used from an OCaml top-level to
               manually step through an interaction with a Neovim instance. Also includes
               a binary for logging communication between a plugin and Neovim.

The vcaml.plugin library

This library provides helpers for two common types of plugins: "oneshot" plugins that model a single call from Neovim to OCaml, and "persistent" plugins that remain alive in the background and field RPC requests. Most of the time you will want to use this library instead of creating a Vcaml.Client.t manually.

The vcaml library

The entrypoint to this library is vcaml.mli. For calling Neovim API functions, you should look at the following modules:

• Buffer - Functions for interacting with buffers
• Window - Functions for interacting with windows
• Tabpage - Functions for interacting with tabs
• Keymap - Functions for interacting with key mappings
• Command - Functions for interacting with commands
• Autocmd - Functions for interacting with auto-commands (logic that runs on events)
• Ui - Functions for attaching and detaching a UI
• Nvim - Miscellaneous functions for interacting with Neovim

In general, the API function nvim_buf_foo will be Buffer.foo; nvim_foo will be Nvim.foo. In some cases functions are renamed or exported differently for clarity. If you are struggling to determine how a function documented in the API is exported in VCaml, you can grep VCaml for the function name, which will appear in the implementation of whichever function(s) export(s) it.

To call Neovim atomically, use Vcaml.block_nvim. Neovim will remain blocked for the duration of the callback, and only commands used with the Client.t provided to the callback will be privileged to run during that time.

To model calls from Neovim to OCaml made with rpcrequest and rpcnotify, see the Ocaml_from_nvim module. If you are using vcaml.plugin, you will not need to call the registration functions yourself.

If you are not using vcaml.plugin, you should use vcaml by first creating a Client.t, then register RPCs you want available with the registration functions in Ocaml_from_nvim, then calling Client.attach, and finally calling Client.close when you are finished with the plugin (or call exit to shut down the program).

Testing with the vcaml.test.helpers library

The vcaml.test.helpers library has useful helper functions for writing VCaml expect tests. They take care of spawning an embedded Neovim instance during the test so you can just write logic with a connected VCaml client. See the plugin tests in plugin/example for example usage.

Debugging VCaml plugins

There are a number of different methods and tools for debugging VCaml programs. Here's a brief guide to the available options.

Print debugging

Printing to Neovim's message history

The simplest way to log messages is to call one of the Neovim API functions for writing output, e.g., Nvim.err_writeln. However, this won't work well in the case where you are trying to debug a blocking request that is not completing, since you won't be able to access the message log while your Neovim instance is blocked.

Printing to stderr

Another option is to write to stderr. Here's a Lua chunk you can use that will consume from stderr and write lines to a log file:

local logfile, error_msg = io.open(path_to_your_log_file, "a")
if not logfile then
  error("Failed to open logfile: " .. error_msg)
end

local opts = {}

function opts.on_stderr(_, data)
  logfile:write(table.concat(data, "\n"))
  logfile:flush()
end

function opts.on_exit(_, exit_code)
  logfile:write("Exited with code " .. exit_code)
  logfile:close()
end

vim.fn.jobstart({ path_to_your_vcaml_exe }, opts)

Note that you'll need to capture stderr to diagnose crashes. Also note that while writing to stdout or stderr work equally well in "persistent" plugins, only stderr can be used to debug "oneshot" plugins since stdio is used for Msgpack RPC.

Debugging communication between Neovim and VCaml

Sometimes print debugging doesn't cut it, and you need to actually observe the traffic between your plugin and Neovim.

Debugging Your Plugin

In the ./debug/bin directory there is a binary for debugging interaction between two Msgpack RPC peers. It has modes for debugging communication both over stdio and over unix domain sockets. You can use this to observe traffic between Neovim and your plugin. See the CLI help for usage information.

Debugging Your Test

If you're trying to write a test for your plugin and the test is not behaving as expected, you can inspect the traffic between your plugin and Neovim by passing ~verbose:true to the test helper function you're using.

Debugging Neovim

You can manually drive Neovim from the OCaml top-level using the vcaml.debug library. Build the top-level for vcaml.debug and follow along below. To get the appropriate value for socket, run :let @+ = v:servername in the target Neovim instance to copy its servername to your clipboard.

# module V = Vcaml_debug.Toplevel_client
module V = Vcaml_debug.Toplevel_client
# #install_printer V.pp (* Improve styling of traffic printing. *)
# let nvim = V.open_ socket
val nvim : V.t = <abstr>
# V.verbose nvim true (* Enable traffic printing for this connection. *)
- : unit = ()
# V.request nvim "nvim_exec2" [ String "echo 'hi'"; Map [] ]
OCaml -> Nvim:
[ 0, 2, "nvim_exec2", [ "echo 'hi'", {} ] ]
- : unit = ()

If you are following these steps manually with a running Neovim instance, at this point you should see "hi" echoed in Neovim. To interpret the messages that are being sent and received, see https://github.com/msgpack-rpc/msgpack-rpc/blob/master/spec.md. The reason the request ID begins at 2 is that vcaml.debug always sends an initial request for the channel.

# V.receive nvim
Nvim -> OCaml: [ 1, 2, nil, {} ]
- : [ `Connection_closed | `Message of Msgpack.t | `Waiting_for_neovim ] =
`Message [ 1, 2, nil, {} ]
# V.request nvim "nvim_exec2" [ String "quit"; Map [] ]
OCaml -> Nvim: [ 0, 3, "nvim_exec2", [ "quit", {} ] ]
- : unit = ()
# V.close nvim
- : unit = ()

If you don't care to see traffic or drive Neovim from OCaml, an often more convenient approach to debugging Neovim is to drive it from a second Neovim instance. Copy the target Neovim's servername as before, and in the instance you're using for debugging, run let socket = sockconnect("pipe", "servername", { "rpc": 1 }) where "servername" is the servername you copied. Now just use rpcrequest and rpcnotify with that socket.