RON is a simple readable data serialization format that looks similar to Rust syntax. It's designed to support all of Serde's data model, so structs, enums, tuples, arrays, generic maps, and primitive values.

{
   "materials": {
        "metal": {
            "reflectivity": 1.0
        },
        "plastic": {
            "reflectivity": 0.5
        }
   },
   "entities": [
        {
            "name": "hero",
            "material": "metal"
        },
        {
            "name": "moster",
            "material": "plastic"
        }
   ]
}

Notice these issues:

1. Struct and maps are the same - random order of exported fields
   • annoying and inconvenient for reading
   • doesn't work well with version control - quoted field names
   • too verbose - no support for enums
2. No trailing comma allowed
3. No comments allowed

Scene( // class name is optional
    materials: { // this is a map
        "metal": (
            reflectivity: 1.0,
        ),
        "plastic": (
            reflectivity: 0.5,
        ),
    },
    entities: [ // this is an array
        (
            name: "hero",
            material: "metal",
        ),
        (
            name: "monster",
            material: "plastic",
        ),
    ],
)

The new format uses (..) brackets for heterogeneous structures (classes), while preserving the {..} for maps, and [..] for homogeneous structures (arrays). This distinction allows to solve the biggest problem with JSON.

Here are the general rules to parse the heterogeneous structures:

There is a very basic, work in progress specification available on the wiki page.

Why not XML?

• too verbose
• unclear how to treat attributes vs contents

Why not YAML?

• significant white-space
• specification is too big

Why not TOML?

• alien syntax
• absolute paths are not scalable

Why not XXX?

• if you know a better format, tell me!

RON is dual-licensed under Apache-2.0 and MIT.