Skip to content

🌽 A simple and pain-free configuration language. Reference libcorn implementation and CLI.

License

Notifications You must be signed in to change notification settings

corn-config/corn

Folders and files

NameName
Last commit message
Last commit date

Latest commit

Β 

History

75 Commits
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 

Repository files navigation

🌽 Corn

A simple and pain-free configuration language.

Corn has been designed using inspiration from JSON and Nix to produce a language that's easy and intuitive to write, good for config files, and has a feature-set small enough you can learn it in minutes. It was born out of the following frustrations:

  • JSON is not a config language, despite how often people use it as one
  • TOML is good for flat structures but gets ugly quickly with deeper objects
  • YAML is far too complex and its whitespace rules make it error-prone

What Corn Is Not

Corn is not a full programming language and does not try to be. There are no variables, there is no interpolation and there are no statement blocks.

Likewise, Corn is not a data exchange format. Unlike JSON or YAML or TOML, you cannot serialize code back into Corn.

Usage

As a binary

Corn can be installed as an executable binary to convert files from the .corn format into any supported output format. Currently, these are:

  • JSON
  • YAML
  • TOML

Windows, Linux and macOS are currently supported.

Install it using cargo:

cargo install corn-cli

Then simply:

corn file.corn-cli
corn file.corn-cli -t yaml

As a library

Corn can be used as a Rust library to deserialize config files directly without needing to convert to other file formats.

crate | docs

The recommended way to do this is using serde into a struct:

use serde::Deserialize;

#[derive(Deserialize)]
struct Config {
    foo: u8
}

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

You can also use libcorn::parse directly to get an AST representation of the config. This can be serialized directly, which offers a faster route when converting to other formats.

A WASM version for use in NodeJS and browsers is also available.

⚠ Note when running under NodeJS you will require --experimental-modules for versions <= 16. On all versions you require --experimental-wasm-modules.

npm

import * as corn from 'libcorn';

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

Writing Corn

This section gives all the outputs in JSON format. Remember you can output in any supported format!

All Corn files must contain a top-level object that contains keys/values.

Keys do not require quotes around them. The first character in the key cannot be whitespace, a number or any of the following characters: . - " $ { [ =. The remaining characters can be any unicode character except whitespace and the following: . =.

Values must be one of the following:

  • String
  • Integer
  • Float
  • Boolean
  • Object
  • Array
  • Null
  • Input

(More on these types below)

A very basic example therefore would be:

{
    hello = "world"
}

Which of course maps to the following in JSON:

{
  "hello": "world"
}

A more complex example would be something like a package.json. This makes use of inputs, various other data types and key chaining:

let {
    $entry = "dist/index.js"
    $author = { name = "John Smith" email = "[email protected]" }
} in {
    name = "example-package"
    version = "1.0.0"
    main = $entry
    bin.executable = $entry
    private = false
    
    author = $author
    author.url = "https://example.com" 
    
    contributors = [ $author ]
    
    scripts.build = "tsc"
    scripts.run = "node dist"
    
    dependencies = {
        dotenv = "^8.2.0"
        // put the rest of your deps here...
    }
    
    devDependencies.typescript = "^4.5"
    
    config.port = 8080
    config.hostname = null
}
This output's a bit longer than the others, so click here to expand it and have a look.
{
  "author": {
    "email": "[email protected]",
    "name": "John Smith",
    "url": "https://example.com"
  },
  "bin": {
    "filebrowser": "dist/index.js"
  },
  "config": {
    "hostname": null,
    "port": 8080
  },
  "contributors": [
    {
      "email": "[email protected]",
      "name": "John smith"
    }
  ],
  "dependencies": {
    "dotenv": "^8.2.0"
  },
  "devDependencies": {
    "typescript": "^4.5"
  },
  "main": "dist/index.js",
  "name": "example-package",
  "private": false,
  "scripts": {
    "build": "tsc",
    "run": "node dist"
  },
  "version": "1.0.0"
}

Types

String

Strings must be surrounded by double quotes. All unicode is supported.

foo = "bar"

Integer

Integers are signed and 64 bit, meaning you can use any value between -9223372036854775808 and 9223372036854775807.

answer = 42

Float

Double precision (64-bit) floats are used.

pi = 3.14159

Boolean

As you'd expect.

not_false = true

Object

Objects use braces to mark the start and end. They contain key/value pairs.

There is no restriction on nesting objects within objects or arrays.

{
  foo = "bar"
  hello = "world"
}

Array

Arrays use square brackets to mark the start and end. Values are space-separated. Like objects, there is no restriction on nesting arrays or objects inside.

    array = [ 1 2 3 ]

You can also include whitespace as you please and mix element types:

{
    array = [ 1 2 3 ]
    array2 = [
        "one"
        2
        true
        { three = 3 }
    ]
}

The above converts to the following JSON:

{
  "array": [
    1,
    2,
    3
  ],
  "array2": [
    "one",
    2,
    true,
    {
      "three": 3
    }
  ]
}

Null

Null values simply use the null keyword:

foo = null

Inputs

Sometimes it may be useful to store some values at the top of the config file, and use or re-use them later on, or even use them to compose more complex values. Corn supports config inputs, akin to variables but they don't change.

All input names start with a dollar sign $ followed by an alphabetic ASCII character or an underscore _. This can then be followed by any amount of alphanumeric ASCII characters or underscores.

Inputs can be used to store any value type, or inside strings.

To declare inputs, you must include a let { } in block at the start of your config.

let { 
    $firstName = "John"
    $lastName = "Smith" 
    
    $birthday = {
        day = 1
        month = 1
        year = 1970
    }
    
} in {
    name = {
        first = $firstName
        last = $lastName
    }
    
    dob = $birthday
}

As well as declaring your own inputs, you can also access any environment variables by prefixing input names with $env_. For example, to use the system PATH:

{
    path = $env_PATH
}

Will output something similar to:

{
  "path": "/home/jake/.cargo/bin:/home/jake/.local/bin:/usr/local/bin:/usr/bin"
}

Environment variable inputs will fall back to regular inputs of the same name, allowing you to create defaults.

Inputs are intentionally quite limited as to what they can do - if you need more power you should use a full language. That said, they hopefully provide a way of quickly viewing/changing values without needing to trawl through the whole file.

Comments

At any point you can start a comment using //. A comment is terminated by a newline \n character. Comments are entirely ignored and not included in the output.

{
    // this is a single-line comment
    foo = bar // this is a mixed-line comment
}

Nesting Keys

Throughout the page, we've created objects within objects using a syntax like this:

{
    foo = {
        bar = "baz"
    }
}

While this is okay for short cases, it can get tiresome very fast and the braces add a lot of noise when reading.

To solve this, keys can be chained to create deep objects instantly:

{
    foo.bar = "baz"
}

Which in JSON is:

{
  "foo": {
    "bar": "baz"
  }
}

You can mix and match chained keys with nested objects at any time:

{
    foo = {
        bar.baz = 42
        qux = true
    }
    
    foo.quux = [ "green eggs" "ham" ]
}

JSON:

{
  "foo": {
    "bar": {
      "baz": 42
    },
    "qux": true,
    "quux": ["green eggs", "ham"]
  }
}

Whitespace

Almost all whitespace in Corn is optional, since keywords and types end as soon as they end. There are only a few exceptions to this:

  • An integer or float following an integer or float must be whitespace separated to tell where one ends and the next starts.
  • References to inputs must terminate with whitespace as otherwise the parser cannot tell where the name ends.
  • There must be whitespace between key=value assignments

This means the below is perfectly valid (although for general consistency and readability this is strongly not recommended):

{
    one={foo="bar" bar="foo"}
    two={foo=1 bar=2}
    three={foo=1.0 bar=2.0}
    four={foo=true bar=false}
    five={foo=null bar=null}
    six={foo={} bar={}}
    seven={foo=[] bar=[]}

    eight=["foo""bar"]
    nine=[truefalse]
    ten=[nullnull]
    eleven=[[][]]
    twelve=[{}{}]
}

And in fact, we could even go as far as to reduce that to a single line:

{one={foo="bar" bar="foo"} two={foo=1 bar=2} three={foo=1.0 bar=2.0} four={foo=true bar=false} five={foo=null bar=null} six={foo={} bar={}} seven={foo=[] bar=[]} eight=["foo""bar"] nine=[truefalse] ten=[nullnull] eleven=[[][]] twelve=[{}{}]}

Editor Support

VSCode

  • Basic syntax highlighting
  • Support Bracket matching and commenting

extension | repo

IntelliJ

  • Syntax highlighting
  • Parsing, lexing and live error checking
  • Support for bracket matching and commenting
  • Basic completion support
  • Basic refactoring support
  • Basic Formatting and code style options

extension | repo

Contributing

Contributions are very welcome, although please do open an issue first as not every potential feature will get merged.

At the moment Corn is in very early stages. If you'd like to help out, I'd absolutely love to see the following:

  • More output formats
  • Improvements and fixes to existing features
  • More tests
  • Better documentation

Running Tests

You must set CORN_TEST=bar as this is required for the environment variable tests.

WASM

WASM support is a feature called wasm which is disabled by default. Make sure to enable it when building:

wasm-pack build -- --features wasm