User guide

This page will walk you through installing Corn as a CLI or library, setting up editor support, and writing your own configuration files.

Code samples with automatically compiled equivalents are included throughout. Each one of these is parsed at build time, meaning you can be sure the code shown works.

Install the CLI

The official Corn CLI can be installed using cargo:

cargo install corn-cli

This installs a binary named corn.

The CLI can take input from .corn files or stdin, and parse it into JSON, YAML or TOML. This allows you to write Corn and use it in existing solutions straight away.

For example:

corn my-file.corn -t yaml

Use corn --help to view all options.

Install a Library

Corn can be integrated into your project with a library.


Add the latest version of libcorn from to your project:

cargo add libcorn

The library supports deserialization via serde:

use serde::Deserialize;

struct Config {
    foo: u8

fn main() {
    let corn = "{foo = 42}";
    let config = corn::from_str::<Config>(corn)?;
    assert_eq!(, 42);

You can also use corn::parse to output an AST in the form of the corn::Value enum. This can be serialized or used to implement any custom logic at a lower level.

fn main() {
    let corn = "{foo = 42}";
    let config: Value = corn::parse(corn)?;

For full details, see the docs. View the crate here.


The libcorn library includes WASM support, allowing you to use the native parser in Node.js and browser environments.

When running under Node.js you will require --experimental-modules for versions <= 16. On all versions you require --experimental-wasm-modules.

Add the latest version of libcorn from NPM to your project:

npm i --save libcorn
# or
yarn add libcorn
# or
pnpm add libcorn

The library exposes a single parse function, which will produce a Map. Parser errors are exposed in the JS environment and can be caught with try/catch.

import * as corn from 'libcorn';

const parsed = corn.parse('{foo = "bar"}');
console.log(parsed); // Map(foo -> "bar")

View the package here.


The libcorn library includes Lua binding support, allowing you to use the native parser.

Currently, Lua requires manual compilation. You must have Lua installed, and compile targeting the installed version.

Clone the repo and build with Lua support:

git clone
cd corn
cargo build -p libcorn --release --features lua54

The following Lua versions are available - change the above feature depending on the target version.

Next copy target/release/ to your Lua project working dir. The library can then be used as follows:

local libcorn = require("libcorn")

local success, res = pcall(libcorn.parse, '{foo = "bar"}')
if success then
    print( -- lua table
    print(res) -- pretty printed error

Editor support

Corn support is available for popular editors.

VS Code

Basic support is available via an extension:

extension | repo


Full support is available via a plugin:

extension | repo


Basic support is available via nvim-treesitter



In circumstances where you want to use Corn but cannot use the CLI or a library, there is a basic web API implementation available.

A publicly available instance is available on, limited to 20 requests per minute and 16k uploads.

The code is also open-source and easily self-hostable.

Full docs can be found in the repo.


The basics

Throughout the rest of this guide, you will see tabbed code blocks. These take the Corn input, parse it using the libcorn JavaScript bindings detailed above, and produce the outputted JSON/YAML/TOML automatically.

Corn is written inside .corn files.

Beginning with the obligatory hello world example:

{ value = "hello world" }

Every Corn file contains a single top-level object. Inside that object, much like JSON, are key/value pairs.

Keys are not quoted and can contain any UTF-8 characters (yes, including emoji…) except whitespace, dots . and equals =. Dots are a special case that we’ll see later on.

Line comments are supported with //.

Values can be any type, again much like JSON:

  a = "hello world" // string
  b = 123 // integer
  c = 3.14 // float
  d = false // boolean
  e = null // null
  f = { } // object
  g = [ ] // array

Note no punctuation is needed between keys.

Corn is very lenient on whitespace and in many places it can be omitted. Keys can all be on one line so long as there is separation between a key and the previous value.

Arrays work in much the same way:

  array = [ "one fish" "two fish" "red fish" "blue fish" ]

Of course, you can mix types including objects in arrays and arrays in objects in any combination to any depth.

  array = [
    "one fish"
    "two fish"
    "red fish"
      complex_fish = true
      name = "Barry"
      age = 2
      eats = [ "other fish" "algae" ]


So far, you’ve seen Corn’s basic syntax. You’re probably thinking it looks pretty close to all the others, so what makes it special? This next section aims to show the cool stuff that answers that.

The functionality available is kept purposely simple - if you need more power than what Corn offers, a configuration language is probably not the right tool for the job. The few tools available to you here though should prove very powerful.

Key chaining

The first and simplest of the features allows you to chain keys with dot.notation to create nested objects instantly. Chains can be as long as you like, and you can mix/match the two syntaxes as you please. Everything is automatically put in the right place.

  foo = {
    qux = "hello world"
  } = 123


You can optionally write a let { } in block before the top-level object. This allows you to declare “input” values, which are effectively constants.

Every input declaration and reference starts with a dollar $, making them easy to spot.

Inputs can be of any value type, and you can use inputs in place of values at any point.

let {
  $myValue = 1234
} in {
  foo = $myValue

You can also reference inputs from other inputs:

let {
  $myValue = 1234
  $myValueArray = [ $myValue ]
} in {
  foo = $myValueArray

You can also ‘environment inputs’, using the $env_ prefix, which read their value from environment variables. These do not need to be declared, although you can to provide a fallback.

let {
    // this is not required if `MY_VALUE` env var is set -
    // no vars can be set in the wasm environment, so the outputs fallback :(
    $env_MY_VALUE = "fallback"
} in {
    my_path = $env_MY_VALUE

Interpolation & Merging

Inputs can be taken a step further, with some basic capabilities to compose more complex values.

With strings, you can interpolate string-type inputs into the values. You can escape interpolation with a backslash \.

let {
  $greeting = "hello"
  $subject = "world"
} in {
  message = "$greeting, $subject"
  escaped = "\$greeting, \$subject"

You can merge object inputs into other objects values, and array inputs into other array values. This allows you to use complex inputs as a ‘base’ where you may need to add or change a few fields.

This is done using the ..$input syntax.

let {
    $baseObj = { a = "hello" }
    $baseArr = [ "one fish" "two fish" ]
} in {
    full = {
        b = "world"
        c = [ ..$baseArr "red fish" "blue fish" ]

You can merge anywhere in an object or array. You can also re-declare keys from merged objects. Since keys and merges are evaluated in order, this allows you to overwrite properties.

Practical example

The following example is lifted from my ironbar config file. It makes heavy use of inputs and their advanced functionalities, as well as some key chaining, to cut down on duplicated content and boilerplate.

let {
    $workspaces = {
        type = "workspaces"
        all_monitors = false
        name_map = {
            1 = "󰙯"
            2 = ""
            3 = ""
            4 = ""
            5 = ""

    $launcher = {
        type = "launcher"
        favorites = [ "firefox" "discord" "steam" ]
        show_names = false
        show_icons = true

    $mpd_base = { type = "music" player_type = "mpd" music_dir = "/home/jake/Music" }
    $mpd_local = { ..$mpd_base name = "music-local" }
    $mpd_server = { ..$mpd_base name = "music-server" host = "media-server:6600" }

    $sys_info = {
        type = "sys_info"
        format = ["{cpu_percent}% " "{memory_percent}% "]
        interval.cpu = 1

    $tray = { type = "tray" }
    $clock = { type = "clock" }

    $clipboard = { type = "clipboard" max_items = 5 truncate.mode = "end" truncate.length = 30 }

    $cmd_shutdown = "shutdown now"
    $cmd_reboot = "reboot"

    $power_btn_base = { type = "button" class="power-btn" }

    $power_menu = {
        type = "custom"
        class = "power-menu"

        bar = [ { type = "button" name="power-btn" label = "" on_click = "popup:toggle" } ]

        popup = [ {
            type = "box"
            orientation = "vertical"
            widgets = [
                { type = "label" name = "header" label = "Power menu" }
                    type = "box"
                    name = "buttons"
                    widgets = [
                        { ..$power_btn_base label = "" on_click = "!$cmd_shutdown" }
                        { ..$power_btn_base label = "" on_click = "!$cmd_reboot" }
        } ]

    $start = [ $workspaces $launcher ]
    $end = [

    $base = {
        start = $start
        end = $end
        icon_theme = "Paper"
} in {
    monitors = {
        DP-1 = { name = "DP-1" ..$base }
        DP-2 = { name = "DP-2" ..$base }

For more information and specifics, check the full specification.